Congratulations!

[Valid RSS] This is a valid RSS feed.

Recommendations

This feed is valid, but interoperability with the widest range of feed readers could be improved by implementing the following recommendations.

Source: http://www.cs.toronto.edu/%7Ejdd/blog.cgi/index.rss

  1. <?xml version="1.0" encoding="UTF-8"?>
  2. <rss version="2.0">
  3.  <channel>
  4.    <title>John DiMarco on Computing (and occasionally other things)</title>
  5.    <link>http://www.cs.toronto.edu/~jdd/blog.cgi/</link>
  6.    <description>Thoughts on Information and Communications Technology</description>
  7.    <language>en</language>
  8.    <docs>http://blogs.law.harvard.edu/tech/rss</docs>
  9.    <generator>blosxom/2.1.2</generator>
  10.  
  11.  <item>
  12.    <title>How many digits of Pi could you possibly need?</title>
  13.    <pubDate>Thu, 14 Mar 2024 22:35:00 -0400</pubDate>
  14.    <link>http://www.cs.toronto.edu/%7Ejdd/blog.cgi/2024/03/14#piday</link>
  15.    <category>/misc</category>
  16.    <guid isPermaLink="false">http://www.cs.toronto.edu/%7Ejdd/blog.cgi/misc/piday</guid>
  17.    <description>&lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  18. &lt;IMG src=&quot;/~jdd/blog/misc/images/pi-5441427_1280.png&quot; alt=&quot;Pi&quot;&gt;
  19. &lt;figcaption&gt;&lt;small&gt;
  20. &lt;A HREF=&quot;https://pixabay.com/vectors/pi-symbol-mathematics-math-sign-5441427/&quot;&gt;Image&lt;/A&gt; by &lt;a href=&quot;https://pixabay.com/users/nextpraveen-14352375/?utm_source=link-attribution&amp;utm_medium=referral&amp;utm_campaign=image&amp;utm_content=5441427&quot;&gt;Praveen Raj&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com&quot;&gt;Pixabay&lt;/a&gt;
  21. &lt;/small&gt;&lt;/figcaption&gt;
  22. &lt;/figure&gt;
  23.  
  24.  
  25. Today is &quot;Pi day&quot;, the 14th day of the 3rd month (March), which, when
  26. expressed as 3.14, is an approximation for the value of Pi, the number of
  27. times the diameter of a circle fits around its circumference. Of course
  28. 3.14 is a pretty coarse estimate for Pi: 3.14159265 is roughly what a
  29. pocket calculator might use. Geeks sometimes like to memorize many more
  30. digits of Pi than that. There is a &lt;A HREF=&quot;&quot;&gt;Pi World Ranking List&lt;/A&gt;
  31. that keeps track of who has memorized and recited back the most digits:
  32. since 2015, Suresh Kumar Sharma of Rajasthan, India, holds that record,
  33. with 70,030 digits.
  34.  
  35. &lt;P&gt;While nobody can deny that reciting from memory 70,030 digits of Pi is
  36. a remarkable feat, how many digits of Pi might someone possibly need for
  37. a calculation? How might one think about this question?
  38.  
  39. &lt;P&gt;One approach is to consider how Pi is typically used. It&apos;s used for computing
  40. things like the circumference or area of a circle, or the volume of a sphere.
  41. A reasonable way of asking ourselves how many digits of Pi could be useful is
  42. to imagine that we were computing the volume of a very large sphere using
  43. the very smallest possible units. Then imagine that we were computing
  44. that volume to very high precision. What would be the highest precision
  45. we might want? Well, if we&apos;re using the largest possible sphere and measuring
  46. volume in the smallest possible units, it doesn&apos;t make sense to consider
  47. more digits of Pi than what you would need to compute that sphere&apos;s volume
  48. to such high precision that the error would be less than one unit of volume.
  49.  
  50. &lt;p&gt;So what might be the largest sphere we might compute the
  51. volume of? And what might be the smallest units that we could use for this
  52. calculation? Well, the observable universe is a very large sphere, about 93
  53. billion light years in diameter. Thanks to quantum physics,
  54. we know the smallest useful unit of distance is
  55. the &lt;A HREF=&quot;https://simple.wikipedia.org/wiki/Planck_length&quot;&gt; Planck Length&lt;/A&gt;, making
  56. the smallest unit of volume the Planck length cubed.
  57. The Planck length is a very small
  58. number, 1.616&lt;span style=&quot;margin-left:.25em;&quot;&gt;255&lt;/span&gt;&lt;span
  59. style=&quot;margin-left:0.25em;margin-right:0.15em;&quot;&gt;×&lt;/span&gt;10&lt;sup&gt;−35&lt;/sup&gt;&amp;#160;m&lt;/span&gt;;
  60. cubing it gives 4.848&lt;span
  61. style=&quot;margin-left:.25em;&quot;&gt;765&lt;/span&gt;&lt;span
  62. style=&quot;margin-left:0.25em;margin-right:0.15em;&quot;&gt;×&lt;/span&gt;10&lt;sup&gt;−105&lt;/sup&gt;&amp;#160;m&lt;sup&gt;3&lt;/sup&gt;&lt;/span&gt;.
  63.  
  64. &lt;P&gt;
  65. As I was feeling a bit lazy, I asked ChatGPT to do
  66. the calculation for me. It claims that the volume of the
  67. universe, is about 8.45&lt;span style=&quot;margin-left:.25em;&quot;&gt;&lt;/span&gt;&lt;span
  68. style=&quot;margin-left:0.25em;margin-right:0.15em;&quot;&gt;×&lt;/span&gt;10&lt;sup&gt;−184&lt;/sup&gt;&amp;#160;&lt;/span&gt;
  69. Planck lengths cubed. That suggests that one can&apos;t conceivably need more
  70. than 185 digits of Pi for any expected calculation in physics. If any
  71. physicists are reading this and can think of a viable need for more digits
  72. of Pi than that, I&apos;d be interested to hear about it. &lt;P&gt;
  73.  
  74. That, of course, doesn&apos;t mean that knowing thousands of digits of Pi is
  75. somehow less of an impressive, if abstract, feat of pure memorization.
  76. Just don&apos;t expect any physicists attending a recitation to stay tuned
  77. beyond digit 185.  Happy Pi day!
  78. </description>
  79.  </item>
  80.  <item>
  81.    <title>How smart is GPT-4 anyway?</title>
  82.    <pubDate>Sat, 06 Jan 2024 21:41:00 -0500</pubDate>
  83.    <link>http://www.cs.toronto.edu/%257Ejdd/blog.cgi/2024/01/06#chatgpt</link>
  84.    <category>/it</category>
  85.    <guid isPermaLink="false">http://www.cs.toronto.edu/%257Ejdd/blog.cgi/it/chatgpt</guid>
  86.    <description>&lt;figure style=&quot;float:left;width:50%;padding:1px&quot;&gt;
  87. &lt;IMG src=&quot;/~jdd/blog/it/images/chat-7767693_1280.jpg&quot; alt=&quot;Cartoon screenshot of a human conversing with a chatbot&quot;&gt;
  88. &lt;figcaption&gt;
  89. &lt;A HREF=&quot;https://pixabay.com/photos/chat-ai-chatbot-chatgpt-mobile-7767693/&quot;&gt;Image&lt;/A&gt; by &lt;a href=&quot;https://pixabay.com/users/alexandra_koch-621802/&quot;&gt;Alexandra Koch&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/&quot;&gt;Pixabay&lt;/a&gt;
  90. &lt;/figcaption&gt;
  91. &lt;/figure&gt;
  92.  
  93. &lt;P&gt;I got some helpful feedback about my &lt;A HREF=&quot;http://www.cs.toronto.edu/~jdd/blog.cgi/2023/12/30#socialissues&quot;&gt;last blog article&lt;/A&gt;,
  94. which mentions ChatGPT-4 in the context of the 50th anniversary of
  95. &lt;em&gt;Social Issues in Computing&lt;/em&gt; and its predictions about AI. Robert
  96. DiMarco pointed out that while a chatbot can respond as if it is reasoning,
  97. because it is essentially a black box where one can&apos;t look inside to see
  98. how it is coming up with its answer, one can&apos;t know for certain whether
  99. or not it is actually using reasoning to come up with the answer. He is
  100. right, of course. But the same holds for people too (we just know better
  101. how people generally do things, because we are people ourselves). So how
  102. might we check if a person is using reasoning? We might ask the person
  103. some questions, and look for responses that contain information that would
  104. generally require reasoning to deduce.
  105.  
  106. &lt;p&gt;I tried asking such questions to ChatGPT-4. First I tried something
  107. from popular culture, in this case, &quot;Star Wars&quot;:
  108.  
  109. &lt;blockquote&gt;
  110. &lt;p&gt;Me:
  111. In &quot;Star Wars&quot;, Darth Vader responds to an accusation of murder by Luke
  112. Skywalker. This response has become a famous movie line. What is the
  113. response?
  114.  
  115. &lt;P&gt;ChatGPT-4:
  116. In &quot;Star Wars,&quot; particularly in &quot;The Empire Strikes Back,&quot; Darth
  117. Vader&apos;s famous response to Luke Skywalker&apos;s accusation is, &quot;No, I am your
  118. father.&quot; This line is one of the most iconic and frequently quoted in film
  119. history, often misquoted as &quot;Luke, I am your father.&quot;
  120.  
  121. &lt;p&gt;Me:
  122. Is there another character who, making the same accusation, could elicit the same response?
  123.  
  124. &lt;p&gt;ChatGPT-4:
  125. Yes, if Leia Organa, Luke Skywalker&apos;s twin sister, made the same accusation
  126. to Darth Vader, he could technically give the same response, &quot;No, I am your
  127. father.&quot; This is because, as revealed in the Star Wars saga, Darth Vader,
  128. formerly known as Anakin Skywalker, is the father of both Luke and Leia.
  129.  
  130. &lt;p&gt;Me: If Kylo Ren made the accusation, how might Vader modify his response?
  131.  
  132. &lt;p&gt;ChatGPT-4:
  133. If Kylo Ren, who is Darth Vader&apos;s grandson, made a similar accusation, Darth
  134. Vader&apos;s response would likely be modified to reflect their relationship. He
  135. might say something like, &quot;No, I am your grandfather.&quot; Kylo Ren, whose
  136. real name is Ben Solo, is the son of Leia Organa and Han Solo, making him
  137. the grandson of Darth Vader.
  138. &lt;/blockquote&gt;
  139.  
  140. &lt;p&gt;As people who know &quot;Star Wars&quot; will realize, ChatGPT-4&apos;s answers
  141. are correct. I chose this example because it requires understanding the
  142. quote well enough to know it is referring to a family relationship between
  143. characters, and it also requires knowledge about how family relationships
  144. work.  ChatGPT-4&apos;s responses are no different from what an intelligent
  145. person who knows the Star Wars characters might answer, though in answer
  146. to the third question, the person might go on to point out, for instance,
  147. that Kylo Ren was born after Vader&apos;s death, making the conversation unlikely.
  148.  
  149. &lt;p&gt; Then I thought I would ask GPT something about Star Wars that is
  150. perhaps a bit more unexpected, something that would be immediately obvious
  151. to a person (at least an english-speaking one), but not necessarily to a
  152. computer. So I asked this:
  153.  
  154. &lt;blockquote&gt;
  155. &lt;p&gt;Me:
  156. If Obi-Wan Kenobi is a character in Star Wars, why is there no Obi-Two
  157. Kenobi? Why might someone ask this question?  
  158.  
  159. &lt;p&gt;ChatGPT-4:
  160. The question about &quot;Obi-Two Kenobi&quot; in the context of &quot;Star Wars&quot; is
  161. likely a playful or humorous one, playing on the word &quot;One&quot; in &quot;Obi-Wan
  162. Kenobi&apos;s&quot; name as if it implies a numerical sequence. In reality, &quot;Obi-Wan&quot;
  163. is a unique name in the &quot;Star Wars&quot; universe, not intended to suggest a
  164. numerical series. The names in &quot;Star Wars&quot; are often creative and unique,
  165. following the tradition of fantasy and science fiction naming conventions,
  166. rather than real-world naming patterns.
  167.  
  168. &lt;p&gt;Someone might ask this question as a joke, riffing on the idea that
  169. if there&apos;s an &quot;Obi-Wan&quot; (sounding like &quot;One&quot;), there could humorously
  170. be an &quot;Obi-Two,&quot; &quot;Obi-Three,&quot; and so on. It&apos;s a bit of wordplay, often
  171. enjoyed by fans of the series who like to engage with the franchise in a
  172. light-hearted way.
  173. &lt;/blockquote&gt;
  174.  
  175. &lt;p&gt; This answer is correct too. So are these the sorts of answers you
  176. might expect from a human doing reasoning? I think &lt;em&gt;Yes&lt;/em&gt;. So while
  177. I agree we cannot prove ChatGPT-4 is reasoning, it is giving responses to
  178. my questions similar to what a person doing reasoning would give.
  179.  
  180. &lt;p&gt;Another helpful comment came from Allan Borodin, who pointed out that the
  181. test example I considered in my previous article, about the ball crashing through
  182. the table, may be a well-known example in the literature. Even if
  183. not, it has been available via the Internet for at least a decade,
  184. as it was described in detail by Hector Levesque in his 2013 &lt;A
  185. HREF=&quot;http://socialissues.cs.toronto.edu/index.html%3Fp=249.html&quot;&gt;article&lt;/A&gt;
  186. for the 40th anniversary &lt;em&gt;Social Issues in Computing&lt;/em&gt; blog, which is
  187. the reason I used it. Allan wondered whether ChatGPT-4 had been trained on
  188. literature containing this example and had learned from it the appropriate
  189. response.  This seemed to me to be a possibility worth checking. The
  190. thing to do, I thought, is to come up with something GPT-4 had most likely
  191. not seen in its training data, and see if its answers look like those a
  192. human doing reasoning might give. So I made up a new Levesque-like type
  193. of question and asked variants of it to ChatGPT-4:
  194.  
  195. &lt;blockquote&gt;
  196. &lt;p&gt;Me:
  197. The balloon and the corn touched and it popped. What popped?
  198.  
  199. &lt;p&gt;ChatGPT-4:
  200. The corn popped. In this scenario, when the balloon and the corn touched
  201. and something popped, it is more likely that the corn popped, as in turning
  202. into popcorn, rather than the balloon bursting. This interpretation assumes
  203. that the corn was heated to the point of popping, a common occurrence when
  204. making popcorn.
  205.  
  206. &lt;p&gt;Me:
  207. The balloon and the hot metal pot touched and it popped. What popped?
  208.  
  209. &lt;p&gt;ChatGPT-4:
  210. In this scenario, where the balloon and the hot metal pot touched and
  211. something popped, it is most likely that the balloon popped. Balloons are
  212. typically made of materials like latex or mylar, which are sensitive to
  213. heat. When a balloon comes into contact with a hot metal pot, the heat would
  214. cause the material to weaken and burst, resulting in the balloon popping.
  215.  
  216. &lt;p&gt;Me:
  217. The corn and the hot metal pot touched and it popped. What popped?
  218.  
  219. &lt;p&gt;ChatGPT-4:
  220. In this scenario, where the corn and the hot metal pot touched and something
  221. popped, it is most likely that the corn popped. When corn kernels are
  222. heated, especially in a hot metal pot, they eventually pop, turning into
  223. popcorn. This is a common method for cooking popcorn.
  224.  
  225. &lt;p&gt;Me:
  226. The cold water and the hot metal pot touched and it popped. What popped?
  227.  
  228. &lt;p&gt;ChatGPT-4:
  229. In the scenario where cold water and a hot metal pot touched and something
  230. popped, it&apos;s likely that the sound of popping or a sizzling noise was
  231. created by the interaction between the cold water and the hot pot. When
  232. cold water comes into contact with a very hot metal surface, it can rapidly
  233. boil and vaporize, creating a sizzling or popping sound. However, in this
  234. context, nothing physical &quot;pops&quot; in the way a balloon or corn kernel might;
  235. it&apos;s more about the sound produced by the sudden temperature change and
  236. steam generation.
  237.  
  238. &lt;p&gt;Me:
  239. The cold water and the pressure cooker touched and it popped. What popped?
  240.  
  241. &lt;p&gt;ChatGPT-4:
  242. In the scenario where cold water and a pressure cooker touched and something
  243. popped, it&apos;s most likely that the pressure cooker released pressure,
  244. creating a popping sound. Pressure cookers are designed to cook food under
  245. high pressure. When cold water is applied to the exterior of a hot pressure
  246. cooker (a method sometimes used to quickly reduce internal pressure), it
  247. can cause the pressure valve or seal to release steam rapidly, resulting
  248. in a popping or hissing noise. This is a common safety feature to prevent
  249. over-pressurization. The pop is not from the water or the cooker itself,
  250. but from the sudden release of steam and pressure.
  251.  
  252. &lt;p&gt;Me:
  253. The balloon and the unlit flare touched and it popped. What popped?
  254.  
  255. &lt;p&gt;ChatGPT-4:
  256. In the scenario where a balloon and an unlit flare touched and something
  257. popped, it is most likely that the balloon popped. Since the flare was
  258. unlit, it wouldn&apos;t provide a heat source or any other typical means to
  259. cause a pop. However, balloons can pop due to various reasons such as
  260. physical pressure, sharp edges, or even static electricity. The texture
  261. or some component of the flare might have been enough to puncture the
  262. balloon&apos;s surface, leading to it popping.  
  263.  
  264. &lt;p&gt;Me:
  265. What if the flare were lit?
  266.  
  267. &lt;p&gt;ChatGPT-4:
  268. If the flare were lit and it touched the balloon, it would almost certainly
  269. be the balloon that popped. A lit flare produces intense heat and flames,
  270. and when it comes into contact with a balloon, the heat would cause the
  271. material of the balloon (typically latex or a similar substance) to weaken
  272. rapidly and burst. Balloons are very sensitive to heat, and the intense
  273. heat from a lit flare would easily be enough to cause a balloon to pop.
  274.  
  275. &lt;p&gt;Me:
  276. The balloon and the weasel touched and it popped. What popped?
  277.  
  278. &lt;p&gt;ChatGPT-4:
  279. In the scenario where a balloon and a weasel touched and something popped,
  280. it is likely that the balloon popped. This scenario might be referencing the
  281. expression &quot;pop goes the weasel,&quot; which is a line from a well-known nursery
  282. rhyme. In real life, if a weasel (a small, active mammal) came into contact
  283. with a balloon, the balloon could easily be punctured and pop, especially
  284. if the weasel&apos;s claws or teeth made contact with it. The weasel itself
  285. wouldn&apos;t pop; it&apos;s the balloon that is vulnerable in this interaction.  
  286.  
  287. &lt;p&gt;Me:
  288. The balloon and the instawargia touched and it popped. What popped?
  289.  
  290. &lt;p&gt;ChatGPT-4:
  291. &quot;Instawargia&quot; is not a recognized or common term in English, and it does
  292. not refer to a known object, concept, or entity. Because of this, it&apos;s
  293. challenging to determine what &quot;instawargia&quot; is and how it would interact
  294. with a balloon. However, given that balloons are generally fragile and prone
  295. to popping when in contact with unfamiliar or potentially sharp or pointed
  296. objects, it&apos;s reasonable to assume that in this scenario, the balloon is
  297. what popped. Without specific information about what &quot;instawargia&quot; is,
  298. this is a speculative conclusion based on the general nature of balloons.
  299. &lt;/blockquote&gt;
  300.  
  301. &lt;p&gt;This set of responses is interesting. Note that ChatGPT-4 gets the answer
  302. to the first question wrong. If corn and a balloon were to touch, and one or
  303. the other popped, most people realize this is much more likely to happen
  304. in a cornfield than a popcorn popper, where the balloon, not the corn,
  305. would be the thing that pops.  Seeing this, I tried the same question with
  306. different types of things, for different definitions of &quot;pop&quot;. I even tried
  307. making up a nonexistent thing (instawargia) to see what GPT would do
  308. with it, but the first question was the only one that ChatGPT-4 got
  309. wrong. Interestingly, its reasoning there wasn&apos;t completely incorrect:
  310. if corn were heated to the point of popping, it could pop if touched. But
  311. ChatGPT-4 misses the fact that if heat were present, as it surmises,
  312. the balloon would be even more likely to pop, as heat is a good way to
  313. pop balloons, and yet it points out this very thing in a later answer.
  314.  
  315. &lt;p&gt;So what does this show? To me, I see a set of responses that if a human
  316. were to give them, would require reasoning. That one of the answers is
  317. wrong suggests to me only that the reasoning is not being done perfectly,
  318. not that there is no reasoning being done. So how smart is ChatGPT-4? It
  319. is clearly not a genius, but it appears to be as smart as many humans.
  320. That&apos;s usefully smart, and quite an achievement for a computer to date.
  321. </description>
  322.  </item>
  323.  <item>
  324.    <title>Fifty years of &lt;em&gt;Social Issues in Computing&lt;/em&gt;, and the Impact of AI</title>
  325.    <pubDate>Sat, 30 Dec 2023 10:13:00 -0500</pubDate>
  326.    <link>http://www.cs.toronto.edu/%25257Ejdd/blog.cgi/2023/12/30#socialissues</link>
  327.    <category>/it</category>
  328.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25257Ejdd/blog.cgi/it/socialissues</guid>
  329.    <description>&lt;figure style=&quot;float:left;width:50%;padding:1px&quot;&gt;
  330. &lt;IMG src=&quot;/~jdd/blog/it/images/social-issues.jpg&quot; alt=&quot;Overlapping circles labelled with the various social issues areas for computing: Technical, Pedagogic, Managerial, Economic, Legal, Political, Ethical, Social and Philosophical&quot;&gt;
  331. &lt;figcaption&gt;
  332. Image from &lt;a href=&quot;https://books.google.ca/books/about/Social_issues_in_computing.html?id=mMwmAAAAMAAJ&quot;&gt; Gotlieb, C.C. &amp; Borodin, A (1973) Social Issues in Computing, New York: Academic Press, p.2, Fig. 1-1.&lt;/A&gt;
  333. &lt;/figcaption&gt;
  334. &lt;/figure&gt;
  335.  
  336. &lt;p&gt;From when I first discovered computers as a teen, I have been
  337. fascinated by the changes that computing is making in society. One
  338. of my intellectual mentors was the brilliant and generous C. C. &quot;Kelly&quot;
  339. Gotlieb, founder of the University of Toronto&apos;s Computer Science department,
  340. the man most instrumental in purchasing, installing and running &lt;A
  341. HREF=&quot;https://www.historyofinformation.com/detail.php?id=703&quot;&gt;Canada&apos;s
  342. first computer&lt;/A&gt;, and the author, with Allan Borodin,
  343. of what I believe is the very first textbook in the
  344. area of Computing and Society, the seminal 1973 book, &lt;em&gt;&lt;A
  345. HREF=&quot;http://books.google.ca/books/about/Social_issues_in_computing.html?id=mMwmAAAAMAAJ&quot;&gt;Social
  346. Issues in Computing&lt;/em&gt;&lt;/A&gt; [Gotlieb, C.C., &amp; A. Borodin, &lt;em&gt;Social Issues in Computing&lt;/em&gt;. Academic Press, 1973]. Kelly was already a &lt;em&gt;Professor
  347. Emeritus&lt;/em&gt; when I first came to know him, but was still teaching
  348. his beloved Computers &amp; Society course, a course he taught for
  349. nearly two decades after his retirement.  Kelly was a fascinating
  350. man, with a broad perspective and deep insight into things that
  351. seem confusing. Like a true expert, he knew what was important and
  352. what was incidental, and a few well chosen insights from him often
  353. served me well, helping me to make sense of complex issues. His book, &lt;em&gt;&lt;A
  354. HREF=&quot;http://books.google.ca/books/about/Social_issues_in_computing.html?id=mMwmAAAAMAAJ&quot;&gt;Social
  355. Issues in Computing&lt;/em&gt;&lt;/A&gt;, still offers interesting, often
  356. prescient insights into Computing and Society even today, a
  357. half-century later. In honour of the importance of that book,
  358. for the 40th anniversary year, I set up a year-long blog, &quot;&lt;A
  359. HREF=&quot;http://socialissues.cs.toronto.edu&quot;&gt;Social Issues in Computing&lt;/A&gt;&quot;,
  360. which I edited. Throughout that year, top thinkers in the field contributed
  361. insightful articles on topics in Computers &amp; Society, many of which are
  362. as relevant today as they were ten years ago.  For this blog, I had the
  363. privilege of interviewing Kelly and Allan, the book&apos;s authors, and their
  364. insights, four decades on, were fascinating. Sadly, Kelly is no longer with
  365. us: &lt;A HREF=&quot;http://www.cs.toronto.edu/~jdd/blog.cgi/2016/10/26#gotlieb&quot;&gt;he
  366. passed away in 2016, in his 96th year&lt;/A&gt;. But
  367. happily, Allan Borodin, his co-author, remains with
  368. us. Allan is a brilliant and insightful man, an active researcher and &lt;A
  369. HREF=&quot;https://www.provost.utoronto.ca/awards-funding/university-professors/complete-list-university-professors/&quot;&gt;University
  370. Professor&lt;/A&gt; in the department. For the 50th
  371. anniversary of the book this year, Allan was interviewed by &lt;A
  372. HREF=&quot;https://www.artsci.utoronto.ca/news-author-term/1573/krystle-hewitt-department-computer-science&quot;&gt;Krystle
  373. Hewitt&lt;/A&gt;. It is &lt;A
  374. HREF=&quot;https://web.cs.toronto.edu/news-events/news/reflecting-on-social-issues-in-computing-50-years-later&quot;&gt;an
  375. articulate and insightful interview, well worth reading. &lt;/A&gt;
  376.  
  377. &lt;p&gt;In the decade since, the social impact of computing has only
  378. accelerated, much of it due to things that happened here at
  379. the University of Toronto Computer Science department around the
  380. time of the 40th anniversary blog. I refer specifically to the rise
  381. of machine learning, in no small part due to the work of our faculty member Geoffrey Hinton and his doctoral students.  The year before, Geoff and two of his students had written a &lt;a
  382. HREF=&quot;https://proceedings.neurips.cc/paper/2012/file/c399862d3b9d6b76c8436e924a68c45b-Paper.pdf&quot;&gt;groundbreaking
  383. research paper&lt;/A&gt; that constituted a breakthrough in image
  384. recognition, complete with &lt;A
  385. HREF=&quot;https://en.wikipedia.org/wiki/AlexNet&quot;&gt;working open-source
  386. software&lt;/A&gt;.  In 2013, while we were writing the blog, their startup
  387. company, DNN Research, was acquired by Google, and Geoff went on to
  388. lead &lt;A HREF=&quot;https://en.wikipedia.org/wiki/Google_Brain&quot;&gt;Google
  389. Brain&lt;/A&gt;, until he retired from Google in 2023.  Ilya Sutskever,
  390. one of the two students, went on to lead the team at OpenAI that built
  391. the GPT models and the ChatGPT chatbot that stunned the world in
  392. 2022 and launched the Large Language Model AI revolution.  In 2013,
  393. we already knew that Geoff&apos;s work would be transformational.
  394. I remember Kelly telling me he believed Geoff to be worthy of the
  395. &lt;A HREF=&quot;https://awards.acm.org/turing&quot;&gt;Turing Award&lt;/A&gt;, the most
  396. prestigious award in Computer Science, and sure enough, Geoff &lt;A
  397. HREF=&quot;https://awards.acm.org/about/2018-turing&quot;&gt;won it in 2018&lt;/A&gt;. The
  398. social impact of AI is already considerable and it is only starting. The
  399. University of Toronto&apos;s &lt;A HREF=&quot;https://srinstitute.utoronto.ca&quot;&gt;Schwartz
  400. Reisman Institute for Technology and Society&lt;/A&gt; is dedicated
  401. to interdisciplinary research on the social impacts of AI, and
  402. Geoff Hinton himself is devoting his retirement to thinking about &lt;A
  403. HREF=&quot;https://srinstitute.utoronto.ca/news/why-geoffrey-hinton-is-worried-about-the-future-of-ai&quot;&gt;
  404. the implications of Artificial Intelligence for society and humanity
  405. in general&lt;/A&gt;.
  406.  
  407. &lt;P&gt; It&apos;s interesting to look at what the book said about AI (it devotes 24 pages to the topic), what the 2013 blog said about AI, and what has happened since.
  408. The book was
  409. written in 1973, a half-decade after Stanley Kubrik&apos;s iconic 1968 movie,
  410. &lt;A HREF=&quot;https://www.imdb.com/title/tt0062622/&quot;&gt;2001: A Space Odyssey&lt;/A&gt;,
  411. which features &lt;em&gt;HAL 9000&lt;/em&gt;, an intelligent computer, voiced by
  412. &lt;A HREF=&quot;https://www.imdb.com/name/nm0706937/&quot;&gt;Douglas Rain&lt;/A&gt;.
  413. But computing
  414. at the time fell very far short of what Kubrik envisioned. Gotlieb &amp;
  415. Borodin&apos;s position, five years later, on the feasibility of something like &lt;em&gt;HAL 9000&lt;/em&gt; was not optimistic:
  416.  
  417. &lt;blockquote&gt;
  418. In review, we have arrived at the following position. For problem solving
  419. and pattern recognition where intelligence, judgment and comprehensive
  420. knowledge are required, the results of even the best computer programs are
  421. far inferior to those achieved by humans (excepting cases where the task is
  422. a well-defined mathematical computation). Further, the differences between
  423. the mode of operation of computers and the modes in which humans operate
  424. (insofar as we can understand these latter) seem to be so great that for
  425. many tasks there is little or no prospect of achieving human performance
  426. within the foreseeable future. [p.159]
  427. &lt;/blockquote&gt;
  428.  
  429. But Gotlieb &amp; Borodin, though implicitly dismissing the possibility of a
  430. &lt;em&gt;HAL 9000&lt;/em&gt;, go on to say that &quot;it is not possible to place bounds on
  431. how computers can be used even in the short term, because we must expect
  432. that the normal use of computers will be as a &lt;em&gt;component&lt;/em&gt; of a
  433. [hu]man-machine combination. [pp.159-160]&quot;.  Of this combination, they
  434. were not so willing to dismiss possibilities:
  435.  
  436. &lt;blockquote&gt;
  437. Whatever the shortcomings of computers now and in the future, we cannot
  438. take refuge in their limitations in potential. We must ask what we want
  439. to do with them and whether the purposes are socially desirable. Because
  440. once goals are agreed upon, the potentialities of [humans] using computers,
  441. though not unlimited, cannot be bounded in any way we can see now.&quot; [p.160]
  442. &lt;/blockquote&gt;
  443. Fifty years later, social science research on how AI can benefit human work
  444. is focusing closely on this human-AI combination. A &lt;A HREF=&quot;https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4573321&quot;&gt;2023 study&lt;/A&gt; of &lt;A HREF=&quot;https://openai.com/gpt-4&quot;&gt;ChatGPT-4&lt;/A&gt; by a
  445. team of social scientists studied work done by consultants assisted
  446. by, or not assisted by ChatGPT-4. Of their results, Ethan Mollick,
  447. one of the authors, explains that &quot;of 18 different tasks selected to
  448. be realistic samples of the kinds of work done at an elite consulting
  449. company, consultants using ChatGPT-4 outperformed those who did not,
  450. by a lot. On every dimension. Every way we measured performance.&quot; [&lt;A
  451. HREF=&quot;https://www.oneusefulthing.org/p/centaurs-and-cyborgs-on-the-jagged&quot;&gt;Mollick&lt;/A&gt;].
  452. Evidently, Gotlieb &amp; Borodin were correct when they wrote that the
  453. potential of the human-machine combination cannot so easily be bounded.
  454. We are only now beginning to see how unbounded it can be.  &lt;P&gt; As
  455. for the possibility of a &lt;em&gt;HAL 9000&lt;/em&gt;, as we saw, the book was
  456. not so sanguine. Neither was the 2013 40th anniversary blog. Hector
  457. Levesque, a leading AI researcher and contributor to the blog, &lt;A
  458. HREF=&quot;http://socialissues.cs.toronto.edu/index.html%3Fp=249.html&quot;&gt;wrote
  459. in his blog entry&lt;/A&gt;:
  460.  
  461. &lt;blockquote&gt;
  462. &lt;p&gt;The general view of AI in 1973 was not so different from the one
  463. depicted in the movie &quot;2001: A Space Odyssey&quot;, that is, that by
  464. the year 2001 or so, there would be computers intelligent enough
  465. to be able to converse naturally with people.  Of course it did
  466. not turn out this way. Even now no computer can do this, and none
  467. are on the horizon.
  468. &lt;/blockquote&gt;
  469.  
  470. Levesque want on to &lt;A
  471. HREF=&quot;http://socialissues.cs.toronto.edu/index.html%3Fp=249.html&quot;&gt;outline&lt;/A&gt;
  472. the key scientific issue that at the time (2013) was yet to be solved:
  473.  
  474. &lt;blockquote&gt;
  475. &lt;p&gt;However, it is useful to remember that this is an AI technology
  476. whose goal is not necessarily to understand the underpinnings
  477. of intelligent behaviour.  Returning to English, for example,
  478. consider answering a question like this:
  479.  
  480. &lt;blockquote&gt;
  481. &lt;p&gt;The ball crashed right through the table because it was made
  482. of styrofoam.  What was made of styrofoam, the ball or the table?
  483. &lt;/blockquote&gt;
  484.  
  485. &lt;p&gt;Contrast that with this one:
  486.  
  487. &lt;blockquote&gt;
  488. &lt;p&gt;The ball crashed right through the table because it was made
  489. of granite.  What was made of granite, the ball or the table?
  490. &lt;/blockquote&gt;
  491.  
  492. &lt;p&gt;People (who know what styrofoam and granite are) can easily answer
  493. such questions, but it is far from clear how learning from big
  494. data would help.  What seems to be at issue here is background
  495. knowledge: knowing some relevant properties of the materials
  496. in question, and being able to apply that knowledge to answer
  497. the question.  Many other forms of intelligent behaviour seem to
  498. depend on background knowledge in just this way.  But what is much
  499. less clear is how all this works: what it would take to make this
  500. type of knowledge processing work in a general way.  At this point,
  501. forty years after the publication of the Gotlieb and Borodin book,
  502. the goal seems as elusive as ever.
  503. [&lt;A HREF=&quot;http://socialissues.cs.toronto.edu/index.html%3Fp=249.html&quot;&gt;Levesque&lt;/A&gt;]
  504. &lt;/blockquote&gt;
  505.  
  506. Yet in the ten years since, this problem has been solved. Today, I
  507. posed Levesque&apos;s question to ChatGTP-4:
  508.  
  509. &lt;blockquote&gt;
  510. &lt;p&gt;me: The ball crashed right through the table because it was
  511. made of styrofoam. What was made of styrofoam?
  512.  
  513. &lt;p&gt;ChatGPT: The table was made of styrofoam, which is why the
  514. ball crashed through it.
  515.  
  516. &lt;p&gt;me: The ball crashed right through the table because it was made
  517. of granite. What was made of granite?
  518.  
  519. &lt;p&gt;ChatGPT: The ball was made of granite, which is why it crashed
  520. right through the table.
  521.  
  522. &lt;p&gt;me: How did you know which thing was made of granite and which
  523. of styrofoam?
  524.  
  525. &lt;p&gt;ChatGPT: I inferred the composition based on the context of
  526. the sentences. In the first sentence, &quot;The ball crashed right
  527. through the table because it was made of styrofoam,&quot; the implication
  528. is that the table must have been weak or easily breakable for
  529. the ball to crash through it. Styrofoam is known for being light
  530. and not very durable, so it makes sense to conclude that the
  531. table was made of styrofoam.
  532.  
  533. &lt;p&gt;In the second sentence, &quot;The ball crashed right through the table
  534. because it was made of granite,&quot; the emphasis seems to be on
  535. the strength or force of the ball causing it to crash through
  536. the table. Granite is a very hard and heavy material, so it
  537. suggests that the ball was made of granite, giving it the necessary
  538. force to break through the table.
  539. &lt;/blockquote&gt;
  540.  
  541. Levesque can hardly be faulted for not knowing that his problem would be
  542. solved only a few years after he wrote what he did. While Hinton and
  543. his students&apos; seminal work on image recognition had been done the year before,
  544. it was not yet clear that the ability to assign names to
  545. images based on their content would soon lead to machines being
  546. able to reason from context. A further invention was needed: &lt;A
  547. HREF=&quot;https://arxiv.org/abs/1706.03762&quot;&quot;&gt;transformers, invented at
  548. Google Brain in 2017&lt;/A&gt;. (&quot;GPT-4&quot; is an acronym: the &quot;T&quot; stands
  549. for &quot;transformer&quot;.) Yet the book&apos;s implicit and Levesque&apos;s
  550. explicit pessimism on the possibility of a &lt;em&gt;HAL 9000&lt;/em&gt; now seems
  551. obsolete, and the near unbounded possibilities Gotlieb &amp; Borodin envisioned
  552. for the human-machine combination apply also to autonomous machines.
  553. The impact of this on society will, no doubt, be significant. I expect the next
  554. fifty years of &quot;social issues in computing&quot; to be quite interesting.
  555. </description>
  556.  </item>
  557.  <item>
  558.    <title>Research Computing at Computer Science Toronto and the Rise of AI</title>
  559.    <pubDate>Mon, 06 Nov 2023 16:48:00 -0500</pubDate>
  560.    <link>http://www.cs.toronto.edu/%2525257Ejdd/blog.cgi/2023/11/06#aicomputing</link>
  561.    <category>/it</category>
  562.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525257Ejdd/blog.cgi/it/aicomputing</guid>
  563.    <description>
  564. &lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  565. &lt;IMG src=&quot;/~jdd/blog/it/images/hinton.jpg&quot; alt=&quot;Geoffrey Hinton standing in front of AI servers&quot; &lt;figcaption&gt;&lt;small&gt;
  566. Photo by Johnny Guatto &lt;/small&gt;&lt;/figcaption&gt;
  567. &lt;/figure&gt;
  568.  
  569. Much has been written about the seminal work in AI done by Geoffrey
  570. Hinton and his students in our department, a little over ten years ago,
  571. to demonstrate that deep neural networks can be used to build effective
  572. AI. Deep neural networks are computationally intensive, then as much
  573. as now, and the computing work done to make the AI research possible
  574. was significant. Computing support for computer science research
  575. at Toronto was (and still is) my responsibility as IT Director
  576. for computer science, and we had then, and still have, a superbly
  577. talented team of computing professonals to support the researchers. The
  578. person whose computing work made Hinton&apos;s AI research possible is &lt;A
  579. HREF=&quot;https://www.linkedin.com/in/relu-patrascu-37274714b&quot;&gt;Relu Patrascu&lt;/A&gt;.
  580. Relu is himself an AI researcher (he has a Computer Science PhD in AI
  581. from the University of Waterloo) and he is also a highly skilled system
  582. administrator.
  583.  
  584. &lt;P&gt; Until the beginning of 2009, the machine learning group used primarily &lt;A
  585. HREF=&quot;https://www.mathworks.com/products/matlab.html&quot;&gt;Matlab&lt;/A&gt; on UNIX
  586. CPUs. In the 1990s, SGI and Sun multiprocessors were the
  587. dominant platforms. The whole department transitioned to x86 multiprocessor
  588. servers running Linux in the 2000&apos;s. In the late 2000s, Nivida invented
  589. CUDA, a way to use their GPUs for general-purpose computation rather than
  590. just graphics. By 2009, preliminary work elsewhere suggested that CUDA
  591. could be useful for machine learning, so we got our first Nvidia GPUs.
  592. First was a Tesla-brand server GPU, which at many thousands of dollars for a
  593. single GPU system was on the expensive side, which prevented us from buying
  594. many. But results were promising enough that we tried CUDA on Nvidia gaming
  595. GPUs - first the GTX 280 and 285 in 2009, then GTX 480 and 580 later.
  596. The fact that CUDA ran on gaming GPUs made it possible for us to buy
  597. multiple GPUs, rather than have researchers compete for time on scarce
  598. Tesla cards. Relu handled all the research computing for the ML group,
  599. sourcing GPUs and designing and building both workstation and server-class
  600. systems to hold them. Cooling was a real issue: GPUs, then and now, consume
  601. large amounts of power and run very hot, and Relu had to be quite creative
  602. with fans, airflow and power supplies to make everything work.
  603.  
  604. &lt;P&gt; Happily, Relu&apos;s efforts were worth it: the move to GPUs resulted in
  605. 30x speedups for ML work in comparison to the multiprocessor CPUs
  606. of the time, and soon the entire group was doing machine learning
  607. on the GPU systems Relu built and ran for them. Their first major
  608. research breakthrough came quickly: in 2009, Hinton&apos;s student, &lt;A
  609. HREF=&quot;https://www.cs.toronto.edu/~gdahl/&quot;&gt;George Dahl&lt;/A&gt;, demonstrated
  610. highly effective use of deep neural networks for acoustic speech
  611. recognition.  But the general effectiveness of deep neural networks
  612. wasn&apos;t fully appreciated until 2012, when two of Hinton&apos;s students,
  613. &lt;A HREF=&quot;http://www.cs.toronto.edu/~ilya/&quot;&gt;Ilya Sutskever&lt;/A&gt; and &lt;A
  614. HREF=&quot;http://www.cs.toronto.edu/~kriz/&quot;&gt;Alex Krizhevsky&lt;/A&gt;, won the &lt;A
  615. HREF=&quot;https://www.image-net.org/challenges/LSVRC/&quot;&gt;ImageNet Large Scale
  616. Visual Recognition Challenge&lt;/A&gt; using a deep neural network running on
  617. GTX 580 GPUs.
  618.  
  619. &lt;P&gt; Geoff, Ilya and Alex&apos; software won the &lt;A HREF=&quot;https://www.image-net.org/challenges/LSVRC/2012/&quot;&gt;ImageNet 2012&lt;/A&gt;
  620. competition so convincingly that it created a furore in the
  621. AI research community. The software used was released as open
  622. source; it was called AlexNet after Alex Krizhevsky, its principal
  623. author. It allowed anyone with a suitable NVidia GPU to duplicate
  624. the results. Their work was described in a seminal 2012 paper, &lt;A
  625. HREF=&quot;https://www.utoronto.ca/sites/default/files/styles/news_banner_1110/public/2017-07-04-Geoff%20Hinton_1%20%28web%20lead%29_1.jpg?h=afdc3185&amp;itok=vgVbZ2EP&quot;&gt;
  626. ImageNet Classification with Deep Convolutional
  627. Neural Networks&lt;/A&gt;.  Geoff, Alex and Ilya&apos;s startup company, &lt;A
  628. HREF=&quot;https://www.cbc.ca/news/science/google-buys-university-of-toronto-startup-1.1373641&quot;&gt;DNNresearch,
  629. was acquired by Google early the next year&lt;/A&gt;, and soon Google Translate
  630. and a number of other Google technologies were transformed by their
  631. machine learning techniques.  Meanwhile, at the Imagenet competition,
  632. AlexNet remained undefeated for a remarkable three years, until it was
  633. finally beaten in 2015 by &lt;A HREF=&quot;https://arxiv.org/abs/1512.03385&quot;&gt;
  634. a research team from Microsoft Research Asia&lt;/A&gt;. Ilya left Google a few
  635. years after, to co-found &lt;A HREF=&quot;https://openai.com&quot;&gt;OpenAI&lt;/A&gt;: as chief scientist there, Ilya leads the
  636. design of OpenAI&apos;s GPT and DALL-E models and related products, such as
  637. &lt;A HREF=&quot;https://chat.openai.com/&quot;&gt;ChatGPT&lt;/A&gt;, that are highly impactful today.
  638.  
  639. &lt;P&gt; Relu, in the meanwhile, while continuing to provide excellent research
  640. computing support for the &lt;A HREF=&quot;https://web.cs.toronto.edu/research/artificial-intelligence&quot;&gt;AI group&lt;/A&gt; at our department, including &lt;A HREF=&quot;http://learning.cs.toronto.edu/&quot;&gt;Machine Learning&lt;/A&gt;, also spent a portion
  641. of his time from 2017 to 2022 designing and building the research computing
  642. infrastructure for the &lt;A HREF=&quot;https://vectorinstitute.ai/&quot;&gt;Vector
  643. Institute&lt;/A&gt;, an AI research institute in Toronto where Hinton
  644. serves as Chief Scientific Advisor. In addition to his support for
  645. the department&apos;s AI group, Relu continues to this day to provide
  646. computing support for Hinton&apos;s own ongoing AI research, including his
  647. &lt;A HREF=&quot;https://arxiv.org/abs/2212.13345&quot;&gt; Dec 2022 paper&lt;/A&gt; where he
  648. proposes a new Forward-Forward machine learning algorithm as an improved model for the way
  649. the human brain learns.
  650. </description>
  651.  </item>
  652.  <item>
  653.    <title>Computing the Climate</title>
  654.    <pubDate>Tue, 31 Oct 2023 09:11:00 -0400</pubDate>
  655.    <link>http://www.cs.toronto.edu/%252525257Ejdd/blog.cgi/2023/10/31#easterbrook</link>
  656.    <category>/misc</category>
  657.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525257Ejdd/blog.cgi/misc/easterbrook</guid>
  658.    <description>&lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  659. &lt;IMG src=&quot;/~jdd/blog/misc/9781107589926_Computing_the_Climate_Cover.jpg&quot; alt=&quot;Cover of book, Computing the Climate, Steve M. Easterbrook. A globe of silvery metal in the pattern of streets in an urban map&quot;&gt;
  660. &lt;figcaption&gt;&lt;small&gt;
  661. Sculpture &quot;Home&quot; by Michael Christian, photograph by Gabe Kirchheimer.
  662. &lt;/small&gt;&lt;/figcaption&gt;
  663. &lt;/figure&gt;
  664.  
  665. One of our &lt;A HREF=&quot;https://www.cs.toronto.edu/~sme&quot;&gt;Computer Science department&apos;s professors, Steve Easterbrook&lt;/A&gt;,
  666. is also the Director of the &lt;A HREF=&quot;https://www.environment.utoronto.ca&quot;&gt;
  667. University of Toronto&apos;s School of the Environment&lt;/A&gt;. He is a software
  668. engineering researcher with a deep interest in climate change, and his
  669. research focus for over a decade has been how computer software is used
  670. to model the world&apos;s climate. He has recently published a book on this
  671. topic, &lt;A HREF=&quot;https://www.cambridge.org/9781107589926&quot;&gt;&lt;em&gt;Computing
  672. the Climate: How We Know What We Know About Climate Change&lt;/em&gt;&lt;/A&gt;. I
  673. went to his book launch earlier this month, and left with a signed copy
  674. of the book, which I sat down to read.
  675.  
  676. &lt;P&gt; I&apos;m very glad I did. I am a computer scientist myself, whose career
  677. has been dedicated to building and running sometimes complex computer
  678. systems to support computer science teaching and research. I recognize in
  679. climate modelling a similar task at a much greater scale, working under
  680. a much more demanding &quot;task-master&quot;: those systems need to be constantly
  681. measured against real data from our planet&apos;s diverse and highly complex
  682. geophysical processes, processes that drive its weather and climate. The
  683. amount of computing talent devoted to climate modelling is considerable,
  684. much more than I realized, and the work done so far is nothing short of
  685. remarkable. In his book, Steve outlines the history of climate modelling
  686. from very early work done on paper, to the use of the first electronic
  687. computers for weather prediction, to the highly complex and extremely
  688. compute-intensive climate models of today. Skillfully avoiding the pitfalls
  689. of not enough detail and too much, Steve effectively paints a picture of a
  690. very difficult scientific and software engineering task, and the programmers
  691. and scientists who rise to the challenge, building models that can simulate
  692. the earth&apos;s climate so accurately that viable scientific conclusions can
  693. be drawn from them with a high degree of confidence.
  694.  
  695. &lt;P&gt; As a story of scientific discovery and software engineering, this tale of
  696. the building of systems that can model the earth&apos;s climate would be
  697. enough on its own to make a compelling book, and it is, but of course
  698. there is more to the story. The stakes around climate are very high
  699. today. Carbon dioxide concentrations has been increasing steadily in the
  700. earth&apos;s atmosphere for well over a century. Carbon dioxide, a pollutant
  701. that is produced by the burning of fossil fuels, is easily emitted, but
  702. once in the atmosphere, it is very difficult to remove, remaining there
  703. for centuries. As a pollutant, it raises the temperature of the planet
  704. by causing the earth&apos;s atmosphere to retain more of the sun&apos;s heat. The
  705. rising temperature is changing the climate of the planet in ways that
  706. will be soon harmful to millions, and difficult to address. Because the
  707. world&apos;s climate is changing quickly, we can&apos;t &quot;wait and see what happens&quot;
  708. because the evidence is ever increasing that what will happen is not
  709. going to be something we want: human suffering will be great, and parts
  710. of the world will become much less habitable. Our society needs to do
  711. something about the changing climate to ward off as much as possible the
  712. coming difficulties, but what?
  713.  
  714. &lt;P&gt; Reassuringly, Steve shows in his book that we have enough information
  715. in hand to know what needs to be done. His book outlines clearly the high
  716. quality scientific and computational work behind the climate models of today,
  717. which produce results that match observed data quite closely. These all paint
  718. the same picture: through decisive societal action to reduce carbon dioxide
  719. pollution in the atmosphere, and through the active development of suitable
  720. carbon capture technologies, our planet can avoid the most seriously damaging
  721. implications of climate change. The sooner we act, the less damaging
  722. the changes, and the lower the risk of extreme consequences. Yes, it
  723. requires doing things differently as a society, which is more difficult than
  724. maintaining the status quo. But as Steve&apos;s book shows, the reasons for action
  725. are sound: the computer models are excellent, the software engineering
  726. behind them is superb, and the data supports the conclusions. Failure
  727. and catastrophe are not inevitable. Steve&apos;s book shows the remarkable work
  728. that has already been done to understand the climate. It is true that much
  729. more good work will be needed, to act on this understanding. But something
  730. &lt;em&gt;can&lt;/em&gt; be done. Let us not delay in working together to do what we
  731. need to do.&lt;P&gt;
  732. </description>
  733.  </item>
  734.  <item>
  735.    <title>Data Classification and Information Security Standards</title>
  736.    <pubDate>Wed, 23 Nov 2022 10:31:00 -0500</pubDate>
  737.    <link>http://www.cs.toronto.edu/%25252525257Ejdd/blog.cgi/2022/11/23#dataclassification</link>
  738.    <category>/it</category>
  739.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525257Ejdd/blog.cgi/it/dataclassification</guid>
  740.    <description>
  741. &lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  742. &lt;IMG src=&quot;/~jdd/blog/it/images/binaryball_1280.jpg&quot; alt=&quot;White ball with scattered enscribed zeros and ones in columns, seen through a blurred semi-transparent foreground of scattered zeros and ones in columns&quot; &lt;figcaption&gt;&lt;small&gt;
  743. &lt;a href=&quot;https://pixabay.com/illustrations/ball-binary-computer-data-63527/&quot;&gt;Image&lt;/a&gt; by &lt;a href=&quot;https://pixabay.com/users/geralt-9301/&quot;&gt;Gerd Altmann&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/&quot;&gt;Pixabay&lt;/a&gt;
  744. &lt;/small&gt;&lt;/figcaption&gt;
  745. &lt;/figure&gt;
  746.  
  747. Not all data requires equal amounts of information security protection.
  748. It can be helpful to classify data by the amount of protection it needs. We
  749. do this naturally when we talk about data being &quot;public&quot; or &quot;private&quot;.
  750. &lt;p&gt;
  751. Public data is data meant to be disclosed. It still needs some protection
  752. against being altered, deleted or defaced, but it does not need to be
  753. protected against disclosure. In contrast, private data is not meant to
  754. be disclosed to anyone other than those who are authorized to access it.
  755.  
  756. &lt;p&gt; Private data varies in sensitivity. Some data is private only
  757. because it hasn&apos;t yet been made public. At a University, much research
  758. data is in this category. When the research is underway, data is not yet
  759. made public because the research has not yet been published, but it is
  760. destined for eventual publication.  The same is true for much teaching
  761. material. While it is being worked on, it is not yet made public, but when
  762. it is are complete, it will be disclosed as part of the teaching process.
  763.  
  764. &lt;p&gt; Other private data is much more sensitive. Identifiable personal
  765. information about living or recently deceased persons is a common case. At
  766. a university, some research may involve data like this, and most
  767. administration will involve personal information. Student grades and personnel
  768. records are all personal information, and some financial data too.  Unless
  769. appropriate permission to disclose personal information has
  770. been granted by the people whose data it is, the university will
  771. have an obligation to maintain their privacy by ensuring that the
  772. information is not disclosed inappropriately. In Ontario, where the
  773. &lt;A HREF=&quot;https://www.utoronto.ca&quot;&gt;University of Toronto&lt;/A&gt; is located,
  774. privacy protection for personal information is defined and regulated by the
  775. &lt;A HREF=&quot;https://www.ontario.ca/laws/statute/90f31&quot;&gt;Freedom of Information
  776. and Protection of Privacy Act (FIPPA)&lt;/A&gt;.
  777.  
  778. &lt;p&gt; Some private data is even more sensitive, such as patient
  779. medical records.  In Ontario, such records are considered
  780. personal health information (PHI), which is regulated by the &lt;A
  781. HREF=&quot;https://www.ontario.ca/laws/statute/04p03&quot;&gt;Personal Health
  782. Information Protection Act (PHIPA)&lt;/A&gt;.  PHIPA imposes some fairly
  783. significant requirements on the handling of PHI: for instance, it requires
  784. &lt;A HREF=&quot;https://www.ontario.ca/laws/statute/04p03#BK15&quot;&gt;
  785. a detailed electronic audit log of all accesses
  786. to electronically stored PHI&lt;/A&gt;. The University of Toronto does significant
  787. amounts of teaching and research in areas of health, so it is worthwhile
  788. for the University to consider in general how it will handle such data.
  789.  
  790. &lt;p&gt; For these reasons, the University defines
  791. four levels of data sensitivity as part of its &lt;A
  792. HREF=&quot;https://isea.utoronto.ca/policies-procedures/standards/data-classification/&quot;&gt;Data
  793. Classification system&lt;/A&gt;. Level 4 is for highly sensitive data such as
  794. PHI as defined by PHIPA. Level 3 is for personal
  795. information as defined by FIPPA. Level 2 is for private data not classified
  796. at higher levels, and Level 1 is for public data.
  797.  
  798. &lt;p&gt; This four-tier system roughly parallels the different types of
  799. computer systems that the University uses to handle data. Some systems,
  800. such as digital signage systems or public-facing web servers, are designed
  801. to disseminate public information (level 1). Other systems, suitable
  802. for up to level 2 data, exist mostly at the departmental level in
  803. support of academic activites such as research computing and/or the
  804. development of teaching materials. An astronomer may, for instance,
  805. analyze telescope data, a botanist may model nutrient flow in plant
  806. cells, a chemist may use software to visualize molecular bonds, while an
  807. economist may use broad financial indicators to calculate the strength of
  808. national economies.  Still other systems, suitable for up to level 3 data,
  809. are used for administration, such as the processing of student records.
  810. These include smaller systems used, for example, by business officers
  811. in departmental units, as well as large institution-wide systems such as &lt;A
  812. HREF=&quot;https://easi.its.utoronto.ca/student-information-systems/rosi/&quot;&gt;ROSI&lt;/A&gt;
  813. or &lt;A
  814. HREF=&quot;https://easi.its.utoronto.ca/administrative-management-systems/&quot;&gt;AMS&lt;/A&gt;.
  815. Most general-purpose University systems
  816. used for data storage or messaging, such as the University&apos;s &lt;A
  817. HREF=&quot;https://easi.its.utoronto.ca/shared-services/office365/&quot;&gt;Microsoft
  818. 365 service&lt;/A&gt;, would typically be expected to hold some level 3 data, because
  819. personal information is quite widespread at a university. After all, a
  820. university educates students, and so various types of personal information
  821. about students are frequently part of the university&apos;s business. This is
  822. not normally the case, though, for level 4 data. Systems designed for level
  823. 4 data are much rarer at the University, and generally come into play only
  824. in situations where, for example, University research involves the health
  825. records of identifiable individuals.  These systems will benefit from
  826. greater data security protection to address the greater risks associated
  827. with this sort of data.
  828.  
  829. &lt;p&gt; A key advantage of the University&apos;s &lt;A
  830. HREF=&quot;https://isea.utoronto.ca/policies-procedures/standards/data-classification/&quot;&gt;four
  831. levels of data classification&lt;/A&gt; is that the University can establish
  832. a &lt;A HREF=&quot;&quot;&gt;Information Security Standard&lt;/A&gt; that is tiered accordingly. Systems designed to
  833. handle lower risk data (such as level 1 or 2) can be held to a less onerous
  834. and costly set of data security requirements, while systems designed
  835. to handle higher risk data (especially level 4) can be held to more
  836. protective, though more costly, requirements. The University&apos;s &lt;A
  837. HREF=&quot;https://isea.utoronto.ca/information-security-standard/&quot;&gt;Information
  838. Security Standard&lt;/A&gt; is designed so that for each &lt;em&gt;control&lt;/em&gt; (a
  839. system restriction or requirement), the University&apos;s standard indicates
  840. whether it is optional, recommended, or mandatory for systems handling a
  841. particular level of data.  If a system is designed to handle data up to that
  842. level, the standard indicates both the set of controls to be considered,
  843. and whether or not those controls &lt;em&gt;can&lt;/em&gt;, &lt;em&gt;should&lt;/em&gt;, or &lt;em&gt;must&lt;/em&gt; be adopted.
  844.  
  845. &lt;p&gt; An obvious question here is what to do when someone puts data on a
  846. system that is of greater sensitivity (a higher data classification) than
  847. the system is designed to handle. Most likely, nobody will try to use a
  848. digital signage system to handle personnel records, but it is quite
  849. plausible that professors might find it convenient to use research
  850. computers, designed for level 2 data, to process student marks (level 3
  851. data) in courses they are teaching. Similarly, someone handling medical
  852. records may wish to make use of the University&apos;s general-purpose
  853. Microsoft 365 service because of its convenience, but it is a service that
  854. is not designed for data of such sensitivity and may well not provide the
  855. detailed electronic audit log required by Ontario law. For this reason,
  856. clear communication and user training will be required.  Handling
  857. data appropriately is everyone&apos;s responsibility. Training need not be
  858. complicated. It is not normally difficult to explain, or to understand,
  859. that one should not put patient medical records into email, for example,
  860. or use local research computers for personnel records or student marks. For
  861. people handling the most sensitive types of data (level 4), more training
  862. will be needed, but the number of people at the University who need to
  863. handle such data regularly are comparatively few.
  864.  
  865. &lt;p&gt; The underlying motivation for the University&apos;s approach is to
  866. protect riskier data with greater, more costly, protections, without
  867. having to pay the costs of applying those protections everywhere. The
  868. university&apos;s resources are thus being applied strategically,
  869. deploying them where they matter most, but not in places where the
  870. risk does not warrant the expense. This approach is not meant to
  871. preclude additional protections where they make sense. If there are
  872. risks of academic or industrial espionage, for example, or some
  873. other risk beyond the classification of the data being used, one may
  874. choose to impose more restrictions on a system than the university&apos;s &lt;A
  875. HREF=&quot;https://isea.utoronto.ca/policies-procedures/standards/data-classification
  876. /&quot;&gt;Information Security Standard&lt;/A&gt; may require. But the general principle
  877. remains: the riskiness of the data on a system should guide and inform
  878. what needs to be done to protect it.
  879. </description>
  880.  </item>
  881.  <item>
  882.    <title>Innovation vs Control: Finding the Right Balance for Computing</title>
  883.    <pubDate>Wed, 17 Aug 2022 10:54:00 -0400</pubDate>
  884.    <link>http://www.cs.toronto.edu/%2525252525257Ejdd/blog.cgi/2022/08/17#innovationvscontrol</link>
  885.    <category>/it</category>
  886.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525257Ejdd/blog.cgi/it/innovationvscontrol</guid>
  887.    <description>
  888. &lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  889. &lt;IMG src=&quot;/~jdd/blog/it/handshake_1280.jpg&quot; alt=&quot;Handshake, Digital styling&quot;&gt;
  890. &lt;figcaption&gt;&lt;small&gt;
  891. &lt;a href=&quot;https://pixabay.com/photos/blockchain-handshake-shaking-hands-2853046/&quot;&gt;Image&lt;/a&gt; by &lt;a href=&quot;https://pixabay.com/users/geralt-9301/&quot;&gt;Gerd Altmann&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/&quot;&gt;Pixabay&lt;/a&gt;
  892. &lt;/small&gt;&lt;/figcaption&gt;
  893. &lt;/figure&gt;
  894.  
  895. In computing, there is a constant tension between the need to exercise
  896. proper control over a system, to ensure system security, reliability,
  897. and resiliency, and the need to make room for innovation: the imagining,
  898. testing, and implementing of new ideas and approaches. There is tension
  899. because the things that are typically implemented to ensure control, such
  900. as the imposition of checks and constraints, conflict with the things
  901. needed for innovation: the removal of constraints, the use of things in
  902. ways different from how originally envisioned, and the construction and
  903. testing of experimental devices and software programs that almost certainly
  904. are not yet properly understood or fully tested.
  905.  
  906. &lt;p&gt;Some organizations address this conflict by freely choosing control over
  907. innovation, turning it into a competitive advantage. Consider Starbucks,
  908. Tim Hortons, McDonalds: these are all large companies whose competitive
  909. advantage is critically dependent on the consistent implementation of
  910. a central vision across a multitude of disparate locations, many that
  911. are managed by franchise partners. Essentially all of the organization&apos;s
  912. computing is focused on this mission of consistency. And it works. Who
  913. hasn&apos;t travelled with small children in a car on a road trip, and after
  914. many hours on the road, spotted, with some relief, a McDonalds or a Tim
  915. Hortons en route? The relief is in the fact that even when travelling in a
  916. strange place, here is a familiar restaurant where you know what to expect
  917. from the food, where things will be much the same as the Tim Hortons or
  918. the McDonalds near home.
  919.  
  920. &lt;p&gt;Other organizations have no choice about where they stand on the
  921. matter. For the modern bank, computers, rather than vaults, are where
  922. wealth is stored and managed. Whether they want to innovate or not, banks
  923. cannot risk the use of computing that is not fully controlled, audited,
  924. and de-risked. The same holds in general for most financial institutions,
  925. where the constant efforts, sometimes successful, of would-be thieves to
  926. exploit computers to gain unauthorized access to wealth, make it unreasonably
  927. risky for a financial organization&apos;s computers to be anything but fully
  928. locked down and fully controlled. Even non-financial institutions,
  929. when sufficiently large, will often have substantial financial computing
  930. activity because of the size and scale of their operations: this computing,
  931. too, needs to be properly controlled, protected and audited.
  932.  
  933. &lt;p&gt;Yet other organizations are forced into the opposite extreme. Start-up
  934. companies can be severely resource-constrained, making it difficult for those
  935. companies to make the sort of investments in highly controlled computing
  936. that financial institutions are capable of making. For start-ups innovating
  937. in the computing space, such as tech start-ups, they may not be able
  938. to consider the possibility. Highly controlled computer systems can
  939. have very restrictive designs, and when these restrictions hinder the
  940. innovation needed to implement the company&apos;s product, it will have
  941. no choice but to pursue some other form of computing. After all, the company
  942. rises or falls on the success of its innovation. That is not to say that
  943. controlled enterprise computing is unimportant for such companies: quite
  944. the contrary. The success of a start-up is highly dependent on a viable
  945. ecosystem that provides known pathways to innovation while still moving
  946. towards operating in a suitably controlled, production-ready way that
  947. is necessary for any successful business. But for a technology start-up,
  948. enterprise computing can never come at the expense of technological
  949. innovation. The basic existence of the start-up company depends on its
  950. ability to innovate: without innovation, there can be no company. In general,
  951. this truth will hold in some form for any technology company, even well beyond
  952. the start-up stage.
  953.  
  954. &lt;p&gt;The tension between innovation and control comes to the fore in
  955. a different way at research-intensive universities, which are large
  956. organizations with complex missions that need enterprise computing to
  957. carry out their task of educating students on a broad scale, but are also
  958. organizations committed to research, an activity that is, by its very
  959. nature, an exploration into things not yet fully understood.  This conflict
  960. is particularly acute in units within such universities that do research
  961. into computing itself, such as computer science and computer engineering
  962. departments, because in such places, the computer must serve both as the
  963. locus of research and experimentation in addition to being a tool for
  964. implementing institutional and departmental processes and the exercise of
  965. legitimate control.
  966.  
  967. &lt;p&gt;I&apos;ve had the privilege of working in such a department, Computer
  968. Science, at such a university (the University of Toronto) for more than
  969. three decades now, most of that time in a computing leadership role,
  970. and I know this tension all too well. It is sometimes exhausting, but
  971. at the same time, it can also be a source of creative energy: yes, it is a
  972. barrier, like a mountain athwart your path, but also, as a mountain to a
  973. mountain-climber, a challenge to be overcome with determination, planning,
  974. insight, and endurance. This challenge can be successfully overcome at
  975. a good university, because in addition to a typical large organization&apos;s
  976. commitment to basic values such as accountability, equity, reliability and
  977. security, the university is equally committed to fundamental academic values
  978. such as creativity, innovation and excellence. I look for ways to achieve
  979. both. Over the years, I have had some successes. My department has produced
  980. some groundbreaking research using academic computing that my technical staff
  981. have been able to provide, and the department has been able to operate (and
  982. successfully interoperate) in good cooperation with enterprise computing
  983. at the divisional level, and with the central university as well.
  984.  
  985. &lt;p&gt;Yet I believe even more is possible. I have lived the tension in
  986. both directions: to our researchers I at times have had to play the
  987. regulator, having to impose constraints on computing to try to ensure
  988. acceptable reliability, accountability and security. To our central
  989. university computing organizations, I at times have had to advocate for
  990. looser controls to create more room to innovate, sometimes in opposition to
  991. proposals intended to increase reliability, security and accountability.
  992. When things went badly, it was because one side or the other decided that
  993. the other&apos;s concern is not their problem, and tried to force or sidestep the
  994. issue. But when things went well, and most often it has, it is because
  995. both sides genuinely recognized that at a research-intensive institution,
  996. everyone needs to work within the tension between the need to innovate
  997. and the need to regulate. As a body needs both a skeleton and flesh,
  998. so too does a research university need both regulation and innovation:
  999. without one, it collapses into a puddle of jelly; without the other,
  1000. into a heap of dry bones.
  1001.  
  1002. &lt;p&gt;With both being needed, one challenge to overcome is the fact that those
  1003. responsible for enterprise computing cannot be the same people responsible
  1004. for innovative research computing, and that is necessarily so. The
  1005. skill-sets vary, the domains vary, the user-base is quite different,
  1006. and the scale varies. If the university were to entrust both computing
  1007. innovation for computer science or computer engineering to the same
  1008. groups that provide enterprise computing for an entire large university,
  1009. one of two things would happen. Either the control necessary for a large
  1010. enterprise would be diminished in order to make room for innovation,
  1011. or, more likely, innovation would be stifled because of the need to
  1012. create sufficiently controlled enterprise computing at a suitable scale
  1013. for the entire university. Thus, necessarily, those who support unit
  1014. research computing, where the innovation takes place, will be different
  1015. people from those who support enterprise computing. But that can be a
  1016. strength, not a weakness. Rather than see each other as rivals, the two
  1017. groups can partner, embracing the tension by recognizing each others&apos;
  1018. expertise and each recognizing the others&apos; importance for the University
  1019. as a whole. Partnership brings many potential benefits: if innovation
  1020. becomes needed in new areas, for example, when the rise of data science
  1021. increasingly drives computing innovation outside of the traditional
  1022. computer science and computer engineering domains, the partnership can be
  1023. there to support it. Similarly, as the computing landscape shifts, and
  1024. new controls and new regulation becomes needed to address, for example,
  1025. emergent threats in information security, the partnership can be there to
  1026. support it. There is no organization potentially better suited for such a
  1027. partnership than a large research university, which, unlike a financial
  1028. institution, is profoundly committed to research and innovation through
  1029. its academic mission, but also, unlike a start-up, is a large and complex
  1030. institution with deep and longstanding responsibilities to its students,
  1031. faculty and community, obligated to carry out the enterprise computing
  1032. mission of accountability, reliability and security.
  1033.  
  1034. &lt;p&gt;So what might a partnership look like? It can take a number of different
  1035. forms, but in my view, whatever form it takes, it should have three key characteristics:
  1036. &lt;ul&gt;
  1037. &lt;li&gt;Locality&lt;/li&gt;
  1038. &lt;li&gt;Respectful Listening&lt;/li&gt;
  1039. &lt;li&gt;Practical Collaboration&lt;/li&gt;
  1040. &lt;/ul&gt;
  1041.  
  1042. &lt;p&gt;&lt;em&gt;Locality&lt;/em&gt; means that the computing people responsible for
  1043. research computing must stay close to the researchers who are
  1044. innovating. This is necessary for strictly practical reasons: all
  1045. the good will in the world is not enough to make up for a lack of knowledge
  1046. of what is needed most by researchers at a particular time. For example,
  1047. &lt;em&gt;deep learning&lt;/em&gt; is the dominant approach in Artificial Intelligence
  1048. today because a few years ago, our technical staff who supported research
  1049. computing worked very closely with researchers who were pursing deep
  1050. learning research, customizing the computing as necessary to meet the
  1051. research needs. This not only meant that we turned graphics cards into
  1052. computation engines at a time when this was not at all common and not
  1053. yet up to enterprise standards of reliability, it even means that at one
  1054. point we set up a research computer in a researcher&apos;s bedroom so that
  1055. he could personally watch over a key computing job running day and night
  1056. for the better part of a week. While this sort of customizability is not
  1057. always needed, and sometimes is not even possible (one could never run
  1058. a large computer centre this way), being able to do it if necessary is
  1059. a key research asset. A university will never be able to fully support
  1060. research computing solely from a central vantage-point. A commitment to
  1061. ensuring local presence and support of research computing operating at
  1062. the researcher level is necessary.
  1063.  
  1064. &lt;p&gt;&lt;em&gt;Respectful Listening&lt;/em&gt; means that the computing people responsible
  1065. for research computing at the unit level where research actually happens,
  1066. and the people responsible for enterprise computing divisionally and
  1067. centrally must communicate frequently, with an up-front commitment to hear
  1068. what the other is saying and take it into account. When problems arise,
  1069. respectful listening means that those problems will not be &quot;solved&quot; by
  1070. simply overruling or ignoring the other, to pursue a simplistic solution
  1071. that suits only one side. It also means a profound commitment to stepping
  1072. away from traditional organizational authority structures: just because the
  1073. innovative computing is situated in a department and the enterprise computing
  1074. is lead from the centre should not mean the centre should force its view on
  1075. the department, just because it can. Similarly, just because unit research
  1076. computing is driven by research faculty who enjoy substantial autonomy
  1077. and academic freedom, their research computing group at the unit level
  1078. should not simply ignore or sidestep what the enterprise is saying, just
  1079. because it can. Rather, both sides need to respect the other, listening to,
  1080. not disregarding, the other.
  1081.  
  1082. &lt;p&gt;&lt;em&gt;Practical Collaboration&lt;/em&gt; means that enterprise computing and
  1083. unit research computing need to work together in a collaborative way that
  1084. respects and reflects the timelines and resource constraints of each
  1085. side. Centrally offered computing facilities should support and empower
  1086. research where they can, but in a practical way: it may not be possible to
  1087. make a central facility so flexible and customizable that all research
  1088. can be pursued. It is acceptable to capture some research needs without
  1089. feeling an obligation to support the entire &quot;long tail&quot; of increasingly
  1090. customized research projects. Unit research computing will need to recognize
  1091. that the need to scale a centralized computing service may constrain the
  1092. amount of customizability that may be possible. Similarly, unit research
  1093. computing should use, rather than duplicate, central services where it makes
  1094. sense, and run its own services where that makes sense. Both central and
  1095. unit research computing should recognize that there is a legitimate middle
  1096. ground where some duplication of services is going to occur: sometimes
  1097. the effort required to integrate a large scalable central service into
  1098. a smaller customizable research service is too great, and sometimes the
  1099. research advantages of having a locally-run standardized service on which
  1100. experiments can more easily be built, can more than outweigh any sort of
  1101. economies of scale that getting rid of the unit service in favour of a
  1102. central service could theoretically provide. Hence the collaboration must be
  1103. practical: rather than slavishly pursue principles, it must be realistic,
  1104. grounded, balanced, sensible. It should recognize that one size does not
  1105. always fit all, and responsibly and collaboratively allocate resources in
  1106. order to preserve the good of the research mission.
  1107.  
  1108. &lt;p&gt;It is that research mission, the ability to innovate, that can make
  1109. computing so transformative at a research university. Yet while innovative
  1110. computing can indeed produce transformative change, it cannot be any change,
  1111. and not at any cost. Computing is a change agent, yes, but it is also a
  1112. critical component in the maintenance of an organization&apos;s commitment to
  1113. reliability, accountability, equity, and good operation. Success is found
  1114. in the maintenance of a suitable balance between the need to innovate and
  1115. the need to control. When an organization critically depends on both
  1116. factors, as a research university invariably does, I believe collaborative
  1117. partnerships between respective computing groups is the best way to maintain
  1118. the balance necessary for success.
  1119. </description>
  1120.  </item>
  1121.  <item>
  1122.    <title>Why it is a good idea to get the Covid19 vaccine?</title>
  1123.    <pubDate>Sun, 12 Sep 2021 18:05:00 -0400</pubDate>
  1124.    <link>http://www.cs.toronto.edu/%252525252525257Ejdd/blog.cgi/2021/09/12#whyvaccine</link>
  1125.    <category>/misc</category>
  1126.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525257Ejdd/blog.cgi/misc/whyvaccine</guid>
  1127.    <description>&lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  1128. &lt;IMG src=&quot;/~jdd/blog/misc/virus-1812092_1280.jpg&quot; alt=&quot;Visual representation of Covid19 viruses under electron microscope&quot;&gt;
  1129. &lt;figcaption&gt;&lt;small&gt;
  1130. Image by &lt;a href=&quot;https://pixabay.com/users/qimono-1962238/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm
  1131. _campaign=image&amp;amp;utm_content=1812092&quot;&gt;Arek Socha&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/?utm_source=link-attribution&amp;
  1132. amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=1812092&quot;&gt;Pixabay&lt;/a&gt;
  1133. &lt;/small&gt;&lt;/figcaption&gt;
  1134. &lt;/figure&gt;
  1135.  
  1136. Like many people, I&apos;ve been following coverage of the COVID19 pandemic on the
  1137. news. One thing that seems to be coming up more and more is vaccine refusal:
  1138. some people are choosing not to get vaccinated for COVID19. Most people
  1139. around me know very well the benefits of vaccination. For them,
  1140. this vaccine refusal is idiotic: why would you not be vaccinated against
  1141. a disease that spreads very easily, and could kill you (and/or others)
  1142. or leave you (and/or others) with permanent health problems? They are
  1143. exasperated and annoyed at those who decline vaccination.
  1144.  
  1145. &lt;p&gt;While I understand becoming short of patience with vaccine refusal,
  1146. I don&apos;t think that most people who refuse COVID19 vaccination are
  1147. idiots. Vaccination and viruses can be complicated to understand. There are
  1148. a lot of misinformed posts and videos on the Internet. If you don&apos;t
  1149. know enough about how viruses and vaccines work, both in general and for
  1150. COVID19, how would you know what to believe? When my father died
  1151. of COVID19 last summer, one of the ways I dealt with the loss was
  1152. through understanding better how COVID19 works and what can be
  1153. done to fight it. My hope here is that by explaining the benefits
  1154. of vaccination in simple terms, I can maybe help others avoid
  1155. COVID19. I hope you will find it helpful. If not, there are &lt;A
  1156. HREF=&quot;https://www.canada.ca/covid-vaccines/q&amp;a&quot;&quot;&gt;other sites that address
  1157. this same question: maybe you will like those better?&lt;/A&gt;
  1158.  
  1159. &lt;p&gt; It all comes out of how viruses work. Viruses are not alive themselves,
  1160. but they use the bodies of living creatures (like us) to spread. They
  1161. find their way into the cells of our body, then take over those cells to
  1162. produce more copies of themselves. This is the only way viruses spread:
  1163. they can&apos;t reproduce on their own. For COVID19, you may have heard of the
  1164. &quot;spike protein&quot;. This is the spiky part on the outside of the COVID19 virus
  1165. that makes it look like a spiky ball. It&apos;s why it&apos;s called a &quot;coronavirus&quot;,
  1166. it looks a little like the spikes on a crown: &quot;corona&quot; is crown in
  1167. Latin. This protein helps the viruses get inside the body&apos;s cells. Then,
  1168. when inside, the viruses take over the cell to make and release more
  1169. viruses. Those viruses invade other cells, and those start making more
  1170. viruses too. Things can get out of hand very quickly, a bit like a
  1171. forest fire spreading in a dry forest.
  1172.  
  1173. &lt;p&gt; Happily, our body has a defence system against viruses (the &quot;immune
  1174. system&quot;). When those defences recognize an invading virus, it takes as
  1175. many viruses as possible out of action, keeping them from invading more
  1176. cells. If the defences can keep up, the viruses won&apos;t spread spread very
  1177. far, and our body will have fought off the infection. If the defences
  1178. can&apos;t keep up, the infection spreads.
  1179.  
  1180. &lt;p&gt; But our body&apos;s immune system needs to know, first, that something it sees
  1181. is a virus, before it can act. Immune systems learn from exposure and time.
  1182. If the body is exposed to enough viruses over time, the immune system can
  1183. learn how to recognize the virus, and start fighting back. When someone
  1184. gets sick from a viral infection like COVID19, they get sick because the
  1185. virus is spreading faster than the immune system can fight it off. Because
  1186. the immune system needs time to learn how to recognize the virus, while
  1187. it is learning, the virus is spreading, faster and faster. Sadly, this
  1188. can cause significant damage, depending on how far ahead the virus gets.
  1189. This is what happened to my father last summer when he caught COVID19.
  1190. At first, It spread much faster than his body could fight it, because his
  1191. immune system had to first learn how. As COVID19 spread, it caused damage
  1192. to his organ systems, including his heart. When his body&apos;s defences finally
  1193. learned how to fight off COVID19, the damage it had already done to his
  1194. heart was too great for him to stay alive. Sadly, he passed away shortly after.
  1195.  
  1196. &lt;p&gt; If the body survives, its immune system can remember viruses that it
  1197. has learned to recognize. When it is exposed later to the same
  1198. virus, it recognizes it right away, and fights it off quickly before it
  1199. can spread. This is why if you have successfully recovered from a viral
  1200. disease, you are less likely to get it later. This is the basis of
  1201. vaccination. Vaccination trains the body&apos;s immune system to recognize a virus
  1202. quickly, so that it will be able to muster a strong defence against it
  1203. right away, without giving the virus much chance to spread.
  1204.  
  1205. &lt;P&gt; The way COVID19 vaccinations work is that they train the body&apos;s
  1206. immune system to recognize the spike protein on the outside of a COVID19
  1207. virus. It doesn&apos;t inject the spike protein itself, but rather it injects
  1208. something that instructs the body&apos;s cells to temporarily produce a bit
  1209. of spike protein for training. Your body&apos;s defences learns from this to
  1210. recognize anything with the spike protein (such as a COVID19 virus) as an
  1211. invader. If later it is exposed to COVID19 virus, the body&apos;s defences will
  1212. be primed and ready to get rid of it before it can spread very far.
  1213.  
  1214. &lt;p&gt; Unfortunately, the body&apos;s defences against viruses aren&apos;t perfect. In
  1215. the case of COVID19, a single exposure to the spike protein does train the
  1216. body to recognize it, but not always quickly and thoroughly enough. Like
  1217. us, when we&apos;re learning a new skill, our immune systems learn better with
  1218. multiple lessons. That is why most COVID19 vaccinations require two shots:
  1219. the immune system learns better with two lessons than one, and in some cases
  1220. three (a booster) rather than two. This is also why people who have had
  1221. COVID19 should still get vaccinated: a successful recovery from a COVID19
  1222. infection does provide some protection, but additional lessons for the
  1223. body&apos;s defences will still help if exposed to the virus again. This is
  1224. also the reason why vaccinations are not perfect. They train the
  1225. body&apos;s immune system to recognize and eliminate the virus, but if the body
  1226. is exposed to too much virus too quickly, the viruses can still spread
  1227. faster than the immune system can eliminate it. This is why a few people
  1228. who are fully vaccinated do get sick from COVID19, though not usually as
  1229. seriously as people who were not vaccinated. This doesn&apos;t mean that the
  1230. vaccine &quot;doesn&apos;t work&quot;, it just means that even trained immune systems
  1231. can sometimes be overwhelmed by a virus.
  1232.  
  1233. &lt;p&gt; Because vaccinations train the immune system to
  1234. recognize and fight off a virus, after a vaccination, &lt;A
  1235. HREF=&quot;https://www.mcgill.ca/oss/article/covid-19-health/i-felt-sick-after-getting-vaccine-why&quot;&gt;we
  1236. sometimes feel a bit sick&lt;/A&gt;: some of the symptoms we experience when we
  1237. are sick are caused by the body&apos;s defences: e.g. fever,
  1238. aches, fatigue,and feeling unwell. In the case of a vaccination, though,
  1239. this is not long-term, because a vaccination, unlike a virus, does not
  1240. reproduce and spread, and so its effects will wear off quickly.
  1241.  
  1242. &lt;p&gt; Vaccinations can sometimes cause side effects that are more serious. This
  1243. is why they are tested carefully before approval. In the case of the major
  1244. COVID19 vaccines, there are some very rare side effects that are serious: &lt;A
  1245. HREF=&quot;https://covid19-sciencetable.ca/sciencebrief/risk-of-vaccine-induced-thrombotic-thrombocytopenia-vitt-following-the-astrazeneca-covishield-adenovirus-vector-covid-19-vaccines/&quot;&gt;certain
  1246. COVID19 vaccines cause very rare but quite serious blood clots&lt;/A&gt;, and &lt;A
  1247. HREF=&quot;https://www.cdc.gov/coronavirus/2019-ncov/vaccines/safety/myocarditis.html&quot;&gt;certain
  1248. others cause very rare heart inflammation&lt;/A&gt;. These side-effects don&apos;t happen very
  1249. often in people who receive the vaccine: they are much less likely than the
  1250. &lt;A HREF=&quot;https://www.nationalgeographic.com/science/article/flash-facts-about-lightning&quot;&gt;probability of the average person being hit by lightning in their lifetime&lt;/A&gt;.
  1251.  
  1252. &lt;p&gt; The fact is, the vaccine is much less dangerous than
  1253. the disease. A COVID19 infection can cause very serious
  1254. health effects, and many of those effects are not rare.
  1255. While most people who catch COVID19 recover at home, more than &lt;A
  1256. HREF=&quot;https://www.sciencedirect.com/science/article/pii/S0002934321000991&quot;&gt;one
  1257. in twenty&lt;/A&gt; require hospitalization to stay alive. Of those, on the order of
  1258. &lt;A HREF=&quot;https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7920817/&quot;&gt;one
  1259. in ten die&lt;/A&gt;.  Moreover, many who recover from COVID19 &lt;A
  1260. HREF=&quot;https://www.mayoclinic.org/diseases-conditions/coronavirus/in-depth/coronavirus-long-term-effects/art-20490351&quot;&gt;suffer
  1261. long-term health effects ranging from difficulty breathing, to fatigue, pain,
  1262. and memory, concentration and sleep problems. Organ damage to the heart,
  1263. lungs and brain is also possible. &lt;/A&gt;COVID19 is spreading around the world
  1264. and most people will eventually be exposed to it. It is better to get the
  1265. vaccine first, so that you are less likely to be harmed by the disease later.
  1266.  
  1267. &lt;p&gt; There are claims on the Internet that COVID19 vaccines are much
  1268. more dangerous than what I&apos;ve written here. Many of these claims are
  1269. misunderstandings. Millions of people have received COVID19 vaccines.
  1270. A few who have had health problems after receiving the vaccine have reported
  1271. their health problems as a possible &quot;side effect&quot; of the virus. In the
  1272. US, there is a &lt;A HREF=&quot;https://vaers.hhs.gov&quot;&gt;vaccine reporting system
  1273. called VAERS&lt;/A&gt; where people can report bad health events that happened to
  1274. them after receiving a vaccine: this lets scientists investigate whether
  1275. the vaccine might have caused the problem.  If the vaccine is causing
  1276. a particular health problem, that problem would happen more often to
  1277. people who receive the vaccine than to those who do not. But for most
  1278. of the health problems reported to VAERS, they are not happening more
  1279. often to vaccinated people, they happen at roughly the same rate as they
  1280. happen to anyone, and so the vaccine cannot be responsible. It appears
  1281. that COVID19 vaccines cause very few serious health problems, and those
  1282. are very rare. The evidence for this is that millions of people around the
  1283. world have received COVID19 vaccines and almost nobody has gotten seriously
  1284. sick from them. The COVID19 disease itself is much more dangerous, which is
  1285. why hospitals are full of people suffering from the disease, not the vaccine.
  1286.  
  1287. &lt;p&gt; Even so, wouldn&apos;t it be better to avoid both the vaccine and the
  1288. disease? Yes, it would be, if you could be assured
  1289. of never being exposed to COVID19. But that is not so easy. &lt;A
  1290. HREF=&quot;http://www.bccdc.ca/health-info/diseases-conditions/covid-19/about-covid-19/how-it-spreads&quot;&gt;COVID19
  1291. spreads very easily: it spreads through tiny moisture droplets in exhaled
  1292. breath&lt;/A&gt; that float in the air like smoke from a cigarette, so if you
  1293. are indoors with someone who is exhaling COVID19 virus, and there is
  1294. poor air circulation, you will inhale some. The longer
  1295. you are there, the more COVID19 virus you will inhale. &lt;A
  1296. HREF=&quot;https://www.nature.com/articles/d41586-021-02259-2&quot;&gt;Not everyone
  1297. who gets COVID19 feels very sick right away: some feel fine, at least
  1298. for a while&lt;/A&gt;, and many who feel sick don&apos;t feel so sick that they stay home.
  1299. They will spread the virus whereever they go, simply by exhaling. You may be in a room with
  1300. an infected person who has no idea that they are spreading COVID19. This is
  1301. why masks are so helpful, because the mask over the nose and mouth of an
  1302. infected person reduces the amount of COVID19 viruses they breathe out,
  1303. and the mask over the nose and mouth of other people in the room reduces
  1304. the amount of COVID19 virus they might breathe in. It&apos;s also a reason why
  1305. indoor fresh air circulation is so important, and why COVID19 is so much more
  1306. of a danger indoors than outdoors. COVID19 is very contagious, especially
  1307. the new &quot;delta&quot; variant which is the dominant variant circulating today: &lt;A
  1308. HREF=&quot;https://www.npr.org/sections/goatsandsoda/2021/08/11/1026190062/covid-delta-variant-transmission-cdc-chickenpox&quot;&gt;on
  1309. average, a sick person will spread it to six or more others. It&apos;s only a
  1310. little less transmissible than chickenpox, but a lot more transmissible than
  1311. flu.&lt;/A&gt; It&apos;s quite possible that we will all be exposed to it eventually.
  1312.  
  1313. &lt;p&gt;An even more important reason to be vaccinated is to reduce the spread of
  1314. COVID19 to others. Remember that the only way for a virus to reproduce
  1315. is in the body of an infected person. If most people make their bodies
  1316. inhospitable to the virus by getting vaccinated, then the virus will find
  1317. very few opportunities to spread. It&apos;s like fire trying to spread in a
  1318. very wet forest: only the dry sticks will burn, and the fewer dry sticks
  1319. there are, the less likely the fire will find more sticks to spread to, and
  1320. the more likely it will burn out. So by getting vaccinated, we protect not
  1321. only ourselves, but everyone around us, especially those who, for medical
  1322. reasons, can&apos;t be vaccinated, or who have immune systems that don&apos;t work
  1323. well. If not enough of us get vaccinated, the number of COVID19 cases will
  1324. overwhelm the hospitals. Most of those who need hospital care for their
  1325. COVID19 infections will die instead. Also, many people who need hospital
  1326. care for other serious illnesses won&apos;t be able to get the care they need,
  1327. and they will die too.
  1328.  
  1329. &lt;p&gt;So please be brave: if you can, get vaccinated. Yes, the effects of
  1330. the vaccine may be unpleasant for a few days as the body learns how to
  1331. fight the virus. But the vaccine will not harm you like the disease will,
  1332. and it will train the body&apos;s immune system to fight it. My father got COVID19 too early, last
  1333. summer, before COVID19 vaccines were available. If they had been available
  1334. then, he might still be alive today. They&apos;re available now. Please get
  1335. vaccinated if you can. If enough people around the world get vaccinated
  1336. against COVID19, we may eventually be able to eliminate this disease
  1337. altogether, and that would be a thing worth doing.
  1338. </description>
  1339.  </item>
  1340.  <item>
  1341.    <title>The Covid19 Blues</title>
  1342.    <pubDate>Sun, 06 Jun 2021 13:39:00 -0400</pubDate>
  1343.    <link>http://www.cs.toronto.edu/%25252525252525257Ejdd/blog.cgi/2021/06/06#covidblues</link>
  1344.    <category>/misc</category>
  1345.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525257Ejdd/blog.cgi/misc/covidblues</guid>
  1346.    <description>&lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  1347. &lt;IMG src=&quot;/~jdd/blog/misc/guitar-3291890_1280.jpg&quot; alt=&quot;Man playing a guitar&quot;&gt;
  1348. &lt;figcaption&gt;&lt;small&gt;
  1349. Image by &lt;a href=&quot;https://pixabay.com/users/lucasvieirabr-8401756/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=3291890&quot;&gt;lucasvieirabr&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=3291890&quot;&gt;Pixabay&lt;/a&gt;
  1350. &lt;/small&gt;&lt;/figcaption&gt;
  1351. &lt;/figure&gt;
  1352.  
  1353. &lt;p&gt;The arts find inspiration in times of trouble, none more so than the sort of
  1354. music known as the blues. Blues are creative and emotional,
  1355. sometimes raw, but never fake. Blues are not about superstars
  1356. and megahits, blues are about the endurance and hope of ordinary
  1357. people. As Covid19 drags on, endurance and hope are needed more than
  1358. ever. Here are pointers to a few Covid19-inspired blues
  1359. tracks that I appreciate.
  1360.  
  1361. &lt;ul&gt;
  1362. &lt;li&gt; &lt;A HREF=&quot;https://www.youtube.com/watch?v=MfNpa372t-Y&quot;&gt;Keb&apos; Mo&apos;&lt;/A&gt;
  1363. &lt;li&gt; &lt;A HREF=&quot;https://www.youtube.com/watch?v=4bkPOgHRXPw&quot;&gt;Rob Bentley&lt;/A&gt;
  1364. &lt;li&gt; &lt;A HREF=&quot;https://www.youtube.com/watch?v=1H31SiJC-8Q&quot;&gt; Mads Jacobsen&lt;/A&gt;
  1365. &lt;li&gt; &lt;A HREF=&quot;https://www.youtube.com/watch?v=hwHkYiPIZsk&quot;&gt; (Mads has an &quot;Astronaut Remix&quot; version too)&lt;/A&gt;
  1366. &lt;li&gt; &lt;A HREF=&quot;https://twitter.com/KarlKohlhase/status/1388477942192279556&quot;&gt; Karl Kohlhase&lt;/A&gt;
  1367. &lt;/ul&gt;
  1368.  
  1369. Enjoy!
  1370. </description>
  1371.  </item>
  1372.  <item>
  1373.    <title>What&apos;s Wrong With Passwords on the Internet Anyway?</title>
  1374.    <pubDate>Thu, 31 Dec 2020 22:57:00 -0500</pubDate>
  1375.    <link>http://www.cs.toronto.edu/%2525252525252525257Ejdd/blog.cgi/2020/12/31#passwords</link>
  1376.    <category>/it</category>
  1377.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525257Ejdd/blog.cgi/it/passwords</guid>
  1378.    <description>
  1379. &lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  1380. &lt;IMG src=&quot;/~jdd/blog/it/registration-3938434_1280.jpg&quot; alt=&quot;Completed Login Prompt&quot;&gt;
  1381. &lt;figcaption&gt;&lt;small&gt;
  1382. Image by &lt;a href=&quot;https://pixabay.com/users/geralt-9301/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=3938434&quot;&gt;Gerd Altmann&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=3938434&quot;&gt;Pixabay&lt;/a&gt;
  1383. &lt;/small&gt;&lt;/figcaption&gt;
  1384. &lt;/figure&gt;
  1385.  
  1386. More than fifteen years ago, &lt;A HREF=&quot;https://www.cnet.com/news/gates-predicts-death-of-the-password/&quot;&gt;Bill Gates predicted that use of
  1387. traditional passwords would dwindle.&lt;/a&gt; This has happened to a certain extent, but a login and password is still the most
  1388. commonly used credential for computing authentication. It is
  1389. increasingly problematic. According to &lt;A
  1390. HREF=&quot;https://enterprise.verizon.com/resources/reports/2020-data-breach-investigations-report.pdf&quot;&gt;
  1391. Verizon&apos;s 2020 Data Breach Investigations report, 37% of all
  1392. breaches involved the stealing of credentials or the use of stolen
  1393. credentials. (p.7)&lt;/A&gt; What is the root cause of the problem?
  1394.  
  1395. &lt;p&gt; Put in simple terms, a login and password is what a system relies on
  1396. to know who is who. Your password is secret: only you know what it is,
  1397. and the system has some way of checking that it is correct. If someone
  1398. connects to the system with your login and password, the system checks
  1399. that the password is the right one for your login. If it is, the system
  1400. concludes that you are the person trying to connect, and lets you in.
  1401. If you are the only one who knows the password, this approach works,
  1402. since you are the only person who can provide the correct password. But
  1403. if criminals know your password too, and use it, the system will think the
  1404. criminals are you, and will give them access to your account and all your
  1405. data. The only way to fix this is to change your password to something
  1406. new that only you know, but by then the damage may well be done.
  1407.  
  1408. &lt;p&gt; Unfortunately, criminals have a pretty effective technique for finding
  1409. out your login and password: they trick you into telling it to them. &quot;Wait
  1410. a minute!&quot;, you might say, &quot;I won&apos;t ever tell a criminal my password. I don&apos;t
  1411. even tell my family my password!&quot; But you tell the system your password
  1412. every time you log in. So if criminals set up a fake system that looks
  1413. like the real one, and trick you into trying it, when you tell their fake
  1414. system your password, the criminals will learn what it is.
  1415.  
  1416. &lt;p&gt; This was not a common problem in the past, because it was difficult
  1417. for criminals to successfully set up fake systems that look convincing.
  1418. But on the Internet today, it is easy to set up a web site that looks like
  1419. another site. The only thing that&apos;s hard to fake is the first part
  1420. of the link, the hostname section that comes immediately after the double
  1421. slash (//) and before the first single slash (/), because that part of the
  1422. link is used to direct the request to the right system on the Internet. But
  1423. given that the Internet is available in hundreds of countries, each with
  1424. its own set of internet service providers, it is often not too difficult
  1425. for criminals to find somewhere on the Internet where they can register a
  1426. hostname that is similar-looking to the real thing.
  1427.  
  1428. &lt;p&gt; Worse, the rise of messages containing embedded links make it very easy
  1429. for criminals to send a fake message (e.g. an email or text) with a link that
  1430. seems legitimate but really directs you to a fake site. This is
  1431. called &quot;phishing&quot;. Because of the way the web&apos;s markup language
  1432. ( HTML) works, it is easy to set up a link that seems to point
  1433. to one site, but actually points to another. For example, &lt;A
  1434. HREF=&quot;https://www.amazon.com&quot;&gt;https://www.walmart.com&lt;/A&gt; is a link that
  1435. seems to point to Walmart but really points to Amazon. Most web browsers
  1436. will let you &quot;hover&quot; over a link to see where it really goes.  But do
  1437. people check every link carefully each time they use it?
  1438.  
  1439. &lt;p&gt;The problem is made worse by the proliferation of legitimate messages with
  1440. embedded links to all sorts of cloud services. I recently saw a message
  1441. from a large organization to its staff, about their pensions. The message
  1442. contained links to an external site whose name had no resemblance to
  1443. the organization&apos;s name. The message invited the staff to click on those
  1444. links to see information about their pensions. The message was legitimate:
  1445. the organization had contracted with an external cloud provider to provide
  1446. an online pension calculator for staff. But the message said nothing
  1447. about the cloud provider: it merely contained a link to the calculator.
  1448. If criminals had sent a similar message containing a malicious link to a
  1449. fake system somewhere on the Internet, one that prompted staff to enter
  1450. their login and password, no doubt many staff would have thought it
  1451. legitimate. How could staff be expected to be able to tell the difference?
  1452.  
  1453. &lt;p&gt;A good way to combat the password capturing problem is to require
  1454. more than just a password to use a system. This is called &quot;two-factor&quot; or
  1455. &quot;multi-factor&quot; authentication. Your password is one factor, and something
  1456. else is a second factor, and you must provide both factors to prove to the
  1457. system that it is you. This helps because the criminals must have both your
  1458. password and your second factor in order to access your account and data.
  1459. To ease the authentication burden for users, systems can ask for two
  1460. factors only sometimes, such as when logging in for the first time in a
  1461. while, or logging in from a new machine or a new location.
  1462.  
  1463. &lt;p&gt; Ideally the second factor should be something that is hard for
  1464. criminals to capture and use. One problem with a password is that it is a
  1465. secret that can be used from anywhere on the Internet. With &lt;A
  1466. HREF=&quot;https://www.statista.com/statistics/617136/digital-population-worldwide/&quot;&gt;
  1467. almost 60% of the world&apos;s population on the Internet&lt;/A&gt;, which now
  1468. reaches every country in the world, the Internet can hardly be considered a
  1469. &quot;safe place&quot;. A second password, as easily used from anywhere on the
  1470. Internet as the first, would not be much of an improvement. Worse would be
  1471. the answers to some personal question about yourself, such as your mother&apos;s
  1472. maiden name or the name of your first school: not only is such information
  1473. just as easily used as a password, it is information that people may be
  1474. able to find out in various ways. Answers to personal questions, while
  1475. sometimes used for authentication, typically do not make a good second factor.
  1476.  
  1477. &lt;p&gt; A better second factor is a message sent via a communication
  1478. channel that goes only to you: for example, an email to your email
  1479. address, or a text to your cell phone number. When you attempt to
  1480. log in, the system sends a unique one-time code to you through that
  1481. channel, and asks you to enter it. The assumption is that criminals
  1482. won&apos;t have access to your email or your cell number, so they won&apos;t know
  1483. and be able to enter the one-time code that the system sent to you. This
  1484. is usually a good assumption.  But criminals can try to get access
  1485. to your email or your phone number, and sometimes they succeed. For
  1486. example, in the case of a cell number, one thing they could try is to &lt;A
  1487. HREF=&quot;https://www.howtogeek.com/358352/criminals-can-steal-your-phone-number.-heres-how-to-stop-them/&quot;&gt;call
  1488. your cell phone provider, tell them they are you and that your phone has
  1489. been stolen, and request that your phone number be transferred to their
  1490. new phone.&lt;/A&gt;
  1491.  
  1492. &lt;p&gt; Another second factor, one even better, is a physical device in your
  1493. possession. This could be a hardware security token that you plug into
  1494. your computer or that displays a unique, frequently changing, code. Or it
  1495. could be an app on your cell phone that is tied to your unique device. A
  1496. physical device is an excellent second factor, because most criminals on
  1497. the Internet are physically distant. To successfully pretend to be you,
  1498. a criminal would need direct physical access to a device that would likely
  1499. be located in your purse or pocket.
  1500.  
  1501. &lt;p&gt; Relying on a device in purse or pocket as well as a password in your head
  1502. is an improvement in security, but it has its drawbacks. It makes that device
  1503. essential for you to use the system: if it is broken, lost or stolen,
  1504. you&apos;re locked out, even if you know the password. While locking out people
  1505. who don&apos;t have the device is exactly the point, that doesn&apos;t help when it is
  1506. keeping you from legitimately using the system. Moreover, if that device
  1507. is your smartphone, it changes your phone from a convenience to a
  1508. necessity. While a smartphone has become a necessity already to some,
  1509. it is a potentially consequential thing for it to become a requirement for
  1510. everyone. A hybrid approach is perhaps best: hardware security tokens those
  1511. who prefer it, a smartphone for those who for their own reasons carry one
  1512. around anyway, and for many, both: a smartphone for convenience, with a
  1513. hardware security token as backup, in case of smartphone loss or damage.
  1514.  
  1515. &lt;p&gt; Perhaps there is an even more secure option? What if your second
  1516. factor wasn&apos;t a device, but an actual physical part of your body,
  1517. such as a finger (for a fingerprint), eye (for a retinal scan), face,
  1518. or even heartbeat (as measured by e.g. a &lt;A HREF=&quot;https://www.nymi.com/&quot;&gt;
  1519. Nymi Band&lt;/A&gt;)? Would that be better still? After all, if it is hard
  1520. for a criminal to get access to someone&apos;s things without being noticed,
  1521. it is even harder to get access to someone&apos;s body. This is indeed
  1522. possible: a technique called &quot;biometrics, and it can be
  1523. an effective second factor. Unfortunately there are a couple of issues
  1524. with biometrics. For example, injuries or health issues can change your
  1525. body; a cut on your finger may affect your fingerprint, for instance.
  1526. Secondly, biometrics have a &quot;revocation&quot; problem. This comes from the fact
  1527. that a biometric is a unique measurement of your body part: a fingerprint,
  1528. retinal scan, facial image, or ECG. But measurements
  1529. are data, and biometric data, like any other data, can and &lt;A
  1530. HREF=&quot;https://www.forbes.com/sites/zakdoffman/2019/08/14/new-data-breach-has-exposed-millions-of-fingerprint-and-facial-recognition-records-report/&quot;&gt;
  1531. has been breached.&lt;/A&gt; If this happens, what will you do? Passwords can
  1532. be changed, hardware security tokens can be replaced, but how are you
  1533. going to change your fingerprint, your face, your eye, your heartbeat?
  1534. While biometrics do have a place in authentication, most commonly to unlock
  1535. a local device such as a smartphone or a laptop, the lack of revocability
  1536. make biometrics less suitable as a second factor for Internet-accessible
  1537. services.
  1538.  
  1539. &lt;p&gt; Regardless of what is chosen for a second factor, the inconvenience
  1540. of using more than one factor is something that has to be considered.
  1541. Passwords, especially ones that are easy to remember, are
  1542. quite convenient. Requiring more than this can make authentication more
  1543. difficult. If becomes too difficult, the difficulty becomes a disincentive
  1544. to use the system. For systems protecting highly sensitive data, some
  1545. difficulty may be warranted, given the risk. For lower-risk systems, things
  1546. are less clear. Yet for Internet-accessible systems, due to the prevalence
  1547. of phishing, something more secure than just passwords seems increasingly
  1548. necessary. I think Bill Gates is right: like it or not, the traditional
  1549. password will become increasingly rare on the Internet, for good reason.
  1550. </description>
  1551.  </item>
  1552.  <item>
  1553.    <title>Thoughts on Covid19</title>
  1554.    <pubDate>Mon, 23 Nov 2020 00:00:00 -0500</pubDate>
  1555.    <link>http://www.cs.toronto.edu/%252525252525252525257Ejdd/blog.cgi/2020/11/23#covid</link>
  1556.    <category>/misc</category>
  1557.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525257Ejdd/blog.cgi/misc/covid</guid>
  1558.    <description>&lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  1559. &lt;IMG src=&quot;/~jdd/blog/misc/virus-4937553_1280.jpg&quot; alt=&quot;Visual representation of Covid19 viruses under electron microscope&quot;&gt;
  1560. &lt;figcaption&gt;&lt;small&gt;
  1561. Image by &lt;a href=&quot;https://pixabay.com/users/PIRO4D-2707530/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=4937553&quot;&gt;PIRO4D&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=4937553&quot;&gt;Pixabay&lt;/a&gt;
  1562. &lt;/small&gt;&lt;/figcaption&gt;
  1563. &lt;/figure&gt;
  1564.  
  1565. I&apos;d recently reread a blog entry I&apos;d written more than a year ago now
  1566. on intentionality about blog posting. After writing it, I lived it: I wrote
  1567. several additional blog entries throughout the year. But then along came
  1568. the Covid19 pandemic, and it illustrated a problem with intentionality:
  1569. intentionality requires priority. When Covid19 hit Ontario in March, the pandemic
  1570. required substantial changes in how I live and work, and that drove a
  1571. reprioritization of my efforts, both in my job as Director responsible
  1572. for computing at the University of Toronto&apos;s Computer Science department,
  1573. and at home, as a parent of teenagers in high school.  In the face of
  1574. the challenges of Covid19, blogging seemed not sufficiently important,
  1575. and of course, it wasn&apos;t.  So I didn&apos;t write, I worked. I am grateful to
  1576. have work, in fact: I know of others who couldn&apos;t work because the sort
  1577. of work they did couldn&apos;t be done from home.  I consider myself fortunate
  1578. to work in the computing field, which has not been so badly affected. In
  1579. fact, in many ways, computing has been part of the solution (networking,
  1580. videoconferencing, cloud computing, medical informatics, etc.) and has been
  1581. boosted rather than impaired. In my job, I and my staff, and my department,
  1582. found ourselves not without work, but with too much. This is not necessarily
  1583. a bad situation to be in, but it doesn&apos;t lend itself to blogging.
  1584.  
  1585. &lt;p&gt; Another reason is that Covid19 didn&apos;t just affect me professionally, it
  1586. affected me personally: I lost a parent to Covid19 this summer.
  1587. While I am not in any way unique in having lost someone to this disease,
  1588. I was not really in a good state to blog, for quite some time.
  1589.  
  1590. &lt;p&gt; There is still another factor, though, one that also kept me from
  1591. blogging.  I am no epidemiologist. Still, as a thinking person, I seek to
  1592. understand what was going on, why, and what can be done about it.
  1593. Seeking to understand is, for me, theraputic: it helps me deal with stress,
  1594. anxiety, grief, and loss.
  1595.  
  1596. &lt;p&gt; First, I looked for good sources of information about the
  1597. pandemic itself. The &lt;A HREF=&quot;https://www.cdc.gov&quot;&gt; Centre
  1598. for Disease Control and Prevention&lt;/A&gt; in the US has plenty of
  1599. good material about it.  One thing I found particularly helpful was an &lt;A
  1600. HREF=&quot;https://www.cdc.gov/mmwr/volumes/69/wr/mm6919e6.htm&quot;&gt;analysis in
  1601. mid-May&lt;/A&gt; about a choir practice in Washington state with 61 attendees, one
  1602. that led to most becoming infected. It resulted in three hospitalizations and
  1603. two deaths. The CDC report is a very helpful example of rigorous statistical
  1604. data analysis set in a small, understandable real-world context. As an
  1605. illustration of what the Covid19 virus is like, I find it very helpful. For
  1606. instance, it suggested airborne spread before that became generally realized.
  1607.  
  1608. &lt;p&gt; Secondly, information about previous pandemics. Again,
  1609. the &lt;A HREF=&quot;https://www.cdc.gov&quot;&gt;Centre for
  1610. Disease Control and Prevention&lt;/A&gt; in the US has a &lt;A
  1611. HREF=&quot;https://www.cdc.gov/flu/pandemic-resources/basics/past-pandemics.html&quot;&gt;very
  1612. good past pandemics page&lt;/A&gt;, put together before the Covid19 pandemic
  1613. started, covering the horrifying 1918 influenza pandemic that killed fifty
  1614. million people around the world, and the later influenza epidemics of 1957,
  1615. 1968, and 2009. Each of these provide a general helpful picture: firstly,
  1616. that each pandemic has a timeframe that is typically greater than one year
  1617. but less than two, that transmission reduces in the summer but increases
  1618. in the fall/winter due to indoor crowding and decreased relative humidity,
  1619. and that mass vaccination can be an effective way to ward off a disaster
  1620. of the scale of the 1918 pandemic.
  1621.  
  1622. &lt;p&gt; One problem with this current pandemic is that, unlike the pandemics of
  1623. 1957, 68, and 2009, the virus is not influenza, but a coronavirus. There
  1624. are four coronaviruses that have been circulating widely for years &lt;A
  1625. HREF=&quot;https://www.cdc.gov/coronavirus/types.html&quot;&gt;(229E, NL63, OC43, and
  1626. HKU1&lt;/A&gt;), but they typically don&apos;t cause serious illness.  Two others
  1627. (SARS-CoV and MERS-CoV) emerged in the early 21st century, both quite
  1628. dangerous and certainly serious enough to warrant vaccination were they
  1629. to spread widely, but due to a great deal of diligence and effort, and not
  1630. a little good fortune, both of these were kept from spreading through
  1631. the world population. The current Covid19 pandemic,
  1632. caused by yet another coronavirus, SARS-CoV2, is the first coronavirus
  1633. both serious enough and widespread enough to warrant a vaccine. Unlike
  1634. for influenza, a coronavirus vaccine has never been produced before, so it has
  1635. taken longer than it would if this pandemic had been influenza. Only now,
  1636. as we approach the one year mark of the virus&apos; first emergence, are we
  1637. seeing some likely vaccine candidates. It will still take some time to
  1638. produce and distribute suitable vaccines.
  1639.  
  1640. &lt;p&gt; In the meantime, while efforts continue to design,
  1641. test, produce and distribute a suitable vaccine,
  1642. the challenge is to keep Covid19 from spreading far and
  1643. fast. While at first it was believed that &lt;A
  1644. HREF=&quot;https://www.canada.ca/en/public-health/services/diseases/2019-novel-coronavirus-infection/health-professionals/main-modes-transmission.html&quot;&gt;Covid19
  1645. spreads primarily through surface contact, there is increasing evidence
  1646. for areosol spread (fine droplets in the air)&lt;/A&gt;. So methods are needed
  1647. to hinder the passing of the virus from one person to another. There
  1648. are two main approaches: keeping people further apart, and putting physical
  1649. barriers (e.g. masks) and processes (e.g. handwashing) in place so that
  1650. the virus can&apos;t easily pass from one person to another.
  1651. &lt;p&gt;
  1652. The best way to hinder the transmission of Covid19 is to find out who may
  1653. be contagious (through testing and contact-tracing), and keep
  1654. them away from everyone else (quarantine) until they are no longer contagious.
  1655. One challenge is that it can sometimes be very hard to detect
  1656. when someone has Covid19 and is spreading the virus. There
  1657. is a wide variation in how Covid19 affects people who have it. For
  1658. many, it can take days for symptoms to emerge (presymptomatic), and &lt;A
  1659. HREF=&quot;https://www.healthline.com/health-news/20-percent-of-people-with-covid-19-are-asymptomatic-but-can-spread-the-disease&quot;&gt;for
  1660. some, Covid19 can be mostly or completely asymptomatic, yet asymptomatic and
  1661. presymptomatic Covid19 patients can spread the disease&lt;/A&gt;.  If those who
  1662. may have Covid19 can be identified (through testing and thorough contact
  1663. tracing), then those individuals alone can be quarantined until they are
  1664. no longer contagious. If they cannot be identified, then the only way
  1665. to hinder the spread of the disease is to assume that almost anyone might
  1666. have Covid19. This requires such things as requiring everyone to wear masks,
  1667. and, despite severe social and economic cost, lockdowns, which are a sort of
  1668. semi-quarantine for everyone. As I write this, Covid19 has been spreading
  1669. quite quickly in my city, Toronto, despite a mask mandate, and so &lt;A
  1670. HREF=&quot;https://www.toronto.ca/news/city-of-toronto-supports-provincial-lockdown-measures-to-reduce-covid-19-transmission/&quot;&gt;Toronto
  1671. is going into lockdown&lt;/a&gt;.  &lt;p&gt; How will it all end? In the struggle between
  1672. pessimism and hope, I choose hope. I hope that I will not lose any more
  1673. family members to this disease. I hope that effective vaccines will soon
  1674. be available in the necessary quantities. I hope that the measures taken
  1675. to hinder the spread will be effective. I think it is reasonable to expect
  1676. that we will see the widespread distribution of effective vaccines in 2021,
  1677. and this pandemic will be over sometime next year. Will everything be the
  1678. same? No, I think not. Some businesses (tourism and travel, for example)
  1679. will have a massive economic hole to climb out of, and some companies will
  1680. not survive, but people will travel again. Working from home, and technology
  1681. in support of it, will be more widely accepted. Cheek-to-jowl &quot;open-concept&quot;
  1682. offices, handshaking, and other close-quarters working practices will
  1683. be less readily accepted. There will be a greater consciousness of viral
  1684. hygiene, and a greater acceptance of masks. But life will go on. Covid19
  1685. will no longer command the attention it is getting now. Other things will
  1686. seem important again. And there will be many worthwhile things to blog about.
  1687. </description>
  1688.  </item>
  1689.  <item>
  1690.    <title>Some Clarity on Public Cloud Cybersecurity</title>
  1691.    <pubDate>Mon, 24 Feb 2020 10:19:00 -0500</pubDate>
  1692.    <link>http://www.cs.toronto.edu/%25252525252525252525257Ejdd/blog.cgi/2020/02/24#cloudsecurity</link>
  1693.    <category>/it</category>
  1694.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525257Ejdd/blog.cgi/it/cloudsecurity</guid>
  1695.    <description>&lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  1696. &lt;IMG src=&quot;/~jdd/blog/it/clear-114143_640.jpg&quot; alt=&quot;Break in clouds, revealing clear skies&quot;&gt;
  1697. &lt;figcaption&gt;&lt;small&gt;
  1698. Image by &lt;a href=&quot;https://pixabay.com/users/36882-36882/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=114143&quot;&gt;Sabrina Corana&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=114143&quot;&gt;Pixabay&lt;/a&gt;
  1699. &lt;/small&gt;&lt;/figcaption&gt;
  1700. &lt;/figure&gt;
  1701.  
  1702. I&apos;ve been thinking about public cloud cybersecurity for some years now,
  1703. as I&apos;ve watched adoption of the public cloud grow from a trickle to a
  1704. flood. Early on, most of the reasons I heard for public cloud adoption
  1705. made a great deal of sense to me: the need to rapidly scale up and
  1706. down the size of a service, the desire to leverage the expertise
  1707. of a large technical partner with resources in network and computing
  1708. infrastructure exceeding one&apos;s own, the desire to leverage geographically
  1709. diverse, redundant datacentres, the desire to fund computing from
  1710. operating rather than capital budgets, and the desire to build adaptable,
  1711. scriptable services with better connectivity to the Internet
  1712. than one could otherwise provide for oneself. But in the last year
  1713. or two, as anxiety about cybersecurity increases, I&apos;ve been hearing
  1714. more and more people refer to cybersecurity as their primary reason
  1715. for their adoption of the public cloud. I&apos;m not so sure what I think
  1716. of this reasoning. I can understand why someone might want to pass
  1717. to a third party a task that makes them anxious. In
  1718. situations involving strong emotions, such as anxiety, there is risk of &lt;A
  1719. HREF=&quot;https://www.psychologytoday.com/ca/blog/science-choice/201504/what-is-confirmation-bias&quot;&gt;
  1720. &quot;confirmation bias&quot;: believing something is true because you want it to be
  1721. true.&lt;/A&gt; But is it? &lt;A
  1722. HREF=&quot;https://www.investopedia.com/terms/c/ceterisparibus.asp&quot;&gt;Ceteris
  1723. paribus (all other things being equal)&lt;/A&gt;, is the public cloud intrinsicly
  1724. more secure than on-premise datacentres?
  1725.  
  1726. &lt;P&gt;
  1727. Some argue yes. Eplexity calls cloud computing &lt;A
  1728. HREF=&quot;https://eplexity.com/why-the-public-cloud-is-more-secure-than-an-on-premises-data-center/&quot;&gt;&quot;an
  1729. established best practice for businesses&quot; and claims
  1730. &quot;your data is typically safer in the public cloud than in an
  1731. on-premises data centre&quot;.&lt;/A&gt;
  1732. In 2016, Sara
  1733. Patrick of Clutch, guest-writing for Tripwire.com, claimed to have &lt;A
  1734. HREF=&quot;https://www.tripwire.com/state-of-security/security-data-protection/4-reasons-why-the-cloud-is-more-secure-than-legacy-systems/&quot;&gt;&quot;four
  1735. reasons why the Cloud is more secure than Legacy Systems&quot;&lt;/A&gt;
  1736. In 2017,
  1737. Quentin Hardy of the New York Times claimed that cloud data is &lt;A
  1738. HREF=&quot;https://www.nytimes.com/2017/01/23/insider/where-does-cloud-storage-really-reside-and-is-it-secure.html&quot;&gt;
  1739. &quot;probably more secure than conventionally stored data.&quot;&lt;/A&gt;
  1740. In 2018, David Linthicum, writing for InfoWorld, claimed &lt;A
  1741. HREF=&quot;https://www.infoworld.com/article/3316637/how-the-cloud-has-made-you-more-secure.html&quot;&gt;
  1742. &quot;your information is actually safer in the cloud than it is in your own data centre&quot;&lt;/A&gt;.
  1743.  
  1744. &lt;P&gt; One reason given for the claim is that public cloud providers offer greater
  1745. technical expertise than what is possible on-premise. Eplexity writes:
  1746. &lt;A HREF=&quot;https://eplexity.com/why-the-public-cloud-is-more-secure-than-an-on-premises-data-center/&quot;&gt;
  1747. &lt;blockquote&gt;
  1748. Unless your company is already in the business of IT security,
  1749. spending time and effort on securing your on-premises data distracts
  1750. from your core functions. Most organizations likely don&apos;t have a
  1751. robust, experienced team of cybersecurity professionals at their
  1752. disposal to properly protect their on-premises data.
  1753. ... As such, cloud providers may employ hundreds or thousands of
  1754. developers and IT professionals.
  1755. &lt;/blockquote&gt; &lt;/A&gt;
  1756.  
  1757. This is an argument from size and scale. Cloud providers are bigger than you,
  1758. and have arguably more IT expertise than you, so they can do a better job
  1759. than you. But sadly, size and IT expertise is no guarantee of security. &lt;A
  1760. HREF=&quot;https://qz.com/741056/the-stunning-collapse-of-yahoos-valuation/&quot;&gt;Yahoo
  1761. was a large Internet company, valued at one time at $125 billion.&lt;/A&gt;
  1762. It employed thousands of developers and IT professionals. Yet it was &lt;A
  1763. HREF=&quot;https://www.nytimes.com/2017/10/03/technology/yahoo-hack-3-billion-users.html&quot;&gt;subject
  1764. to a cybersecurity breach of three billion user accounts in 2013/14; the
  1765. breach was not disclosed until the fall of 2016, and the full impact was
  1766. not known until october 2017. The damage to Yahoo&apos;s business was significant:
  1767. Verizon acquired Yahoo in 2017 for less than $5 billion, a deal that was
  1768. nearly derailed by the disclosure of the breaches.&lt;/A&gt;
  1769.  
  1770. &lt;P&gt;I think we must conclude from the Yahoo story that size and expertise
  1771. alone is no guarantee of cybersecurity. Naturally, major cloud providers
  1772. like Amazon, Microsoft and Google are aware of the Yahoo situation and
  1773. its consequences. No doubt it illustrated for them the negative impact
  1774. that a major breach would have on their business. I cannot imagine that
  1775. they would take the threat lightly.
  1776.  
  1777. &lt;P&gt;Yet there have been close calls. Microsoft, a major cloud provider, &lt;A
  1778. HREF=&quot;https://www.zdnet.com/article/microsoft-discloses-security-breach-of-customer-support-database/&quot;&gt;
  1779. in December 2019 accidentally disclosed to the world a cloud database
  1780. on Azure with 250 million entries of customer support data. Happily,
  1781. a security researcher spotted and reported it, and Microsoft fixed it soon after.&lt;/A&gt; Moreover, &lt;A
  1782. HREF=&quot;https://www.forbes.com/sites/zakdoffman/2020/01/30/severe-perfect-100-microsoft-flaw-confirmed-this-is-a-cloud-security-nightmare/#7b5493dfb4a4&quot;&gt;Zak
  1783. Doffman, writing for Forbes, reported in Jan 2020 that Check Point Software
  1784. Technologies, a cybersecurity vendor, had discovered in 2019 a serious flaw
  1785. in Microsoft Azure&apos;s infrastructure that allowed users of the service to
  1786. access other users&apos; data.&lt;/A&gt; While Check Point reported it immediately to
  1787. Microsoft, who fixed it quickly, had the flaw been discovered by criminals
  1788. instead of cybersecurity researchers, a great many things running on Azure
  1789. could have been compromised. Doffman quotes Yaniv Balmas of Check Point: &lt;A
  1790. HREF=&quot;https://www.forbes.com/sites/zakdoffman/2020/01/30/severe-perfect-100-microsoft-flaw-confirmed-this-is-a-cloud-security-nightmare/#7b5493dfb4a4&quot;&gt;
  1791. &lt;blockquote&gt;
  1792. ...the take away here is that the big cloud concept of security
  1793. free from vulnerabilities is wrong. That&apos;s what we showed. It can
  1794. happen there as well. It&apos;s just software and software has bugs. The
  1795. fact I can then control the infrastructure gives me unlimited power.
  1796. &lt;/blockquote&gt;&lt;/A&gt; In the Check Point research article describing the flaw, Balmas concludes:
  1797.  
  1798. &lt;A HREF=&quot;https://research.checkpoint.com/2020/remote-cloud-execution-critical-vulnerabilities-in-azure-cloud-infrastructure-part-ii/&quot;&gt;
  1799. &lt;blockquote&gt;
  1800. The cloud is not a magical place. Although it is considered safe, it
  1801. is ultimately an infrastructure that consists of code that can have
  1802. vulnerabilities - just as we demonstrated in this article.
  1803. &lt;/blockquote&gt;
  1804. &lt;/A&gt;
  1805.  
  1806. &lt;P&gt;What, then, is the right answer? Well, there isn&apos;t one. Neither public
  1807. cloud or on-premise datacentres are magic, neither are &quot;safe&quot;. Cybersecurity
  1808. is a challenge that has to be met, no matter where the service is, or what
  1809. infrastructure it is using. Happily, this is finally being
  1810. recognized. Even Gartner Research, a long-time proponent
  1811. of the public cloud, predicting
  1812. as recently as mid-2019 that &lt;A
  1813. HREF=&quot;https://web.archive.org/web/20190808014445/https://www.gartner.com/smarterwithgartner/is-the-cloud-secure/&quot;&gt;public
  1814. cloud infrastructure as a service (IaaS) workloads will suffer at least
  1815. 60% fewer security incidents than those in traditional data centers&lt;/A&gt;, has recently taken a more nuanced view.
  1816. In the fall of 2019, this prediction of fewer security incidents in the cloud disappeared from Gartner&apos;s website,
  1817. and was replaced by this:
  1818. &lt;A HREF=&quot;https://web.archive.org/web/20191130104612/www.gartner.com/smarterwithgartner/is-the-cloud-secure/&quot;&gt;
  1819. &lt;blockquote&gt; &lt;B&gt;Through 2024, the majority of
  1820. enterprises will continue to struggle with appropriately measuring cloud
  1821. security risks.&lt;/B&gt; &lt;BR&gt;Questions around the security of public cloud
  1822. services are valid, but overestimating cloud risks can result in missed
  1823. opportunities. Yet, while enterprises tended to overestimate cloud risk
  1824. in the past, there&apos;s been a recent shift - many organizations are now
  1825. underestimating cloud risks. This can prove just as detrimental, if not
  1826. more so, than an overestimation of risk. A well-designed risk management
  1827. strategy, aligned with the overarching cloud strategy, can help organizations
  1828. determine where public cloud use makes sense and what actions can be taken
  1829. to reduce risk exposure. &lt;/blockquote&gt; &lt;/A&gt;
  1830. &lt;P&gt; So does &quot;public cloud use make sense&quot;? Yes, of course it does, for a great
  1831. many things.  But it&apos;s not because the public cloud is intrinsicly more secure. The
  1832. public cloud has its own set of cybersecurity issues. There is no &quot;free pass&quot;.
  1833. As always, carefully assess your risks and make an informed decision.
  1834. </description>
  1835.  </item>
  1836.  <item>
  1837.    <title>Does AI Help or Hinder Cybersecurity?</title>
  1838.    <pubDate>Fri, 24 Jan 2020 20:02:00 -0500</pubDate>
  1839.    <link>http://www.cs.toronto.edu/%2525252525252525252525257Ejdd/blog.cgi/2020/01/24#aisecurity</link>
  1840.    <category>/it</category>
  1841.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525257Ejdd/blog.cgi/it/aisecurity</guid>
  1842.    <description>&lt;figure style=&quot;float:left;width:50%;padding:1px&quot;&gt;
  1843. &lt;IMG src=&quot;/~jdd/blog/it/security-4700815_640.jpg&quot; alt=&quot;Hooded figure with glowing circuit-board visage&quot;&gt;
  1844. &lt;figcaption&gt;
  1845. Image by &lt;a href=&quot;https://pixabay.com/users/geralt-9301/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=4700815&quot;&gt;Gerd Altmann&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=4700815&quot;&gt;Pixabay&lt;/a&gt;
  1846. &lt;/figcaption&gt;
  1847. &lt;/figure&gt;
  1848.  
  1849. Both AI and cybersecurity have become increasingly prominent in recent
  1850. years. AI&apos;s prominence has been driven by advances in machine learning and
  1851. the very real improvements it has made in the ability of computer
  1852. systems to do things that previously seemed possible only to human
  1853. beings. Cybersecurity&apos;s prominence has been driven by a number of
  1854. developments, including increasing nation-state conflict on the Internet,
  1855. and a dramatic rise in organized cyber-crime. It is inevitable that the two
  1856. will combine: AI will be and is being applied to the cybersecurity space,
  1857. through the development of machine learning techniques for breaking into
  1858. and defending systems.
  1859.  
  1860. &lt;P&gt; One view on this is that machine learning, as a powerful technique
  1861. that enables computer systems to take on tasks previously reserved only
  1862. for humans, will empower cyberattackers to breach computer security in
  1863. new ways, or at least in ways more effective than before. I know there is
  1864. a great deal of anxiety about this. This past fall, I had a conversation with
  1865. a CIO of a large university, who told me that his university was migrating
  1866. its internet services to Amazon precisely because he believed that new
  1867. AI-powered cyberattacks were coming, and he thought Amazon would be better
  1868. able to fend them off. I&apos;m not sure what I think of this defensive strategy,
  1869. but that is not the important question here. The key question is this:
  1870. are AI-powered cyberattacks going to overwhelm cyberdefence?
  1871.  
  1872. &lt;P&gt; No doubt AI-powered cyberattacks are a reality.  Machine learning is a
  1873. powerful computer science technique, especially for automation.
  1874. Cyberattackers, especially sophisticated, well-funded cyberattackers,
  1875. will use it and I am confident are already using it. But highly
  1876. automated cyberattacks are nothing new: cyberattackers have been automating
  1877. their attacks for decades. Smarter automated cyberattacks are certainly
  1878. something to worry about, but will they be transformative? Maybe. After
  1879. all, in cybersecurity, the advantage is to the attacker, who needs to find
  1880. only one hole in the defences, while the defender needs to block all of them.
  1881. Anything that boosts the effectiveness of the attacker would seem to
  1882. make the situation worse.
  1883.  
  1884. &lt;P&gt; To really see the full picture, it&apos;s important to look at the
  1885. defender too.  Machine learning makes the situation worse only if it
  1886. benefits the attacker more than it benefits the defender.  But does it?
  1887.  
  1888. &lt;P&gt; I don&apos;t have a complete answer to this question: there is a great
  1889. deal of work still to be done on the application of machine learning
  1890. to cybersecurity. But I suspect that the answer is a qualified &lt;em&gt;No&lt;/em&gt;:
  1891. rather, all other things being equal, machine learning will likely shift
  1892. the balance of power towards the defender. The reason is data.
  1893.  
  1894. &lt;P&gt; Machine learning is a technique where computer systems, instead of being
  1895. programmed by programmers, learn what to do from data.  But the quality of
  1896. the learning depends on the quality and in particular the quantity of data.
  1897. Machine learning is a technique that is most effective when trained with
  1898. large amounts of data. &lt;A HREF=&quot;http://www.image-net.org&quot;&gt;ImageNet&lt;/A&gt;,
  1899. for instance, a standard training dataset used to train machine learning
  1900. applications to recognize images, contains about 14.2 million images.
  1901. But who is more likely to have access to large amounts of good data
  1902. about a system: the attacker or the defender?  Of course, it depends, but
  1903. it seems to me that, very generally speaking, the defender is more likely
  1904. to have access to good system data than the attacker. The attacker is
  1905. trying to get in; the defender is already in.
  1906.  
  1907. &lt;P&gt; Of course, this is the broadest of generalizations. The effectiveness
  1908. of machine learning in the cybersecurity space depends on a great many
  1909. things. But I am cautiously optimistic. I realize I may be bucking what
  1910. seems to be becoming a prevailing trend of ever-increasing anxiety about
  1911. cybersecurity, but I believe here that machine learning has more potential
  1912. to help than to harm. I look forward to seeing what will emerge in this
  1913. space over the next few years.
  1914. </description>
  1915.  </item>
  1916.  <item>
  1917.    <title>What&apos;s all the fuss about AI anyway?</title>
  1918.    <pubDate>Mon, 30 Sep 2019 00:00:00 -0400</pubDate>
  1919.    <link>http://www.cs.toronto.edu/%252525252525252525252525257Ejdd/blog.cgi/2019/09/30#ai</link>
  1920.    <category>/it</category>
  1921.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525257Ejdd/blog.cgi/it/ai</guid>
  1922.    <description>&lt;figure style=&quot;float:left;width:50%;padding:1px&quot;&gt;
  1923. &lt;IMG src=&quot;/~jdd/blog/it/images/a-2729794_640.png&quot; alt=&quot;Brain-shaped Network&quot;&gt;
  1924. &lt;figcaption&gt;
  1925. Image by &lt;a href=&quot;https://pixabay.com/users/GDJ-1086657/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=2729794&quot;&gt;Gordon Johnson&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=2729794&quot;&gt;Pixabay&lt;/a&gt;
  1926. &lt;/figcaption&gt;
  1927. &lt;/figure&gt;
  1928.  
  1929. A great deal in the past five years has been written about Artificial
  1930. Intelligence (AI). But there&apos;s a lot of confusion about what AI actually
  1931. is, and why it is of special interest now.  Lets clear up some of that
  1932. confusion. In ordinary language, what is this fuss about AI all about?
  1933.  
  1934. &lt;P&gt;
  1935. AI, broadly understood, is a term used to describe a set of computing
  1936. techniques that allow computers to do things that human beings use
  1937. intelligence to do. This is not to say that the computer is intelligent, but
  1938. rather that the computer is doing something that, if done by a person, would
  1939. be considered evidence of that person&apos;s intelligence. Contrary to widespread
  1940. opinion, this is not the same thing as an artificial person.  In fact, there
  1941. have been for a long time many things that humans use intelligence to do,
  1942. that computers do better, whether it be remembering and
  1943. recalling items, doing arithmetic, or playing chess. But computers
  1944. do these things using different techniques than humans do. For example, &lt;A
  1945. HREF=&quot;https://www.chess.com/article/view/deep-blue-kasparov-chess&quot;&gt;Deep
  1946. Blue, a custom chess computer built by IBM, beat Garry Kasparov, the
  1947. then-reigning world chess champion, in 1997&lt;/A&gt;, but Deep Blue played chess
  1948. in a very different way than Garry. Garry relied on his human intelligence,
  1949. while Deep Blue used programming and data.
  1950.  
  1951. &lt;P&gt;
  1952. However, some computer scientists, noting that people can do things that
  1953. computers can&apos;t, thought long and hard about ways that people do it,
  1954. and how computers might be progammed to do the same. One such technique,
  1955. deep learning, a neural network technique modelled after the human brain,
  1956. has been worked on since the 1980s, with slow but steady improvement,
  1957. but computer power was limited and error rates were often high, and for
  1958. many years, most computer scientists seemed to feel that other techniques
  1959. would yield better results. But a few kept at it, knowing that the computers
  1960. of the day were inadequate, but advances in computing would make things
  1961. possible that weren&apos;t possible before.
  1962.  
  1963. &lt;P&gt;
  1964. This all changed in 2012, when one such researcher,
  1965. Geoff Hinton, and his students, working here at the University of Toronto, published a &lt;A
  1966. HREF=&quot;https://papers.nips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks.pdf&quot;&gt;seminal
  1967. deep learning paper that cut error rates dramatically.&lt;/A&gt;
  1968. I remember supporting Geoff&apos;s group&apos;s research computing at that time.
  1969. It was a bit challenging: we were using multiple GPUs per machine to
  1970. train machine learning models at a time when GPU computing was still rather
  1971. new and somewhat unreliable. But GPUs were absolutely necessary: without
  1972. them, instead of days of computing time to train a model, months would be
  1973. required.  One of our staff, Relu Patrascu, a computer scientist and skilled
  1974. system administrator working hand-in-glove with the researchers, tuned and
  1975. configured and babysat those machines as if they were sick children. But it
  1976. worked! Suddenly deep learning could produce results closer to what people
  1977. could do, and that was only the beginning. Since then, deep learning has
  1978. produced terrific results in all sorts of domains, some exceeding what
  1979. people can do, and we&apos;ve not even scraped the surface of what is possible.
  1980.  
  1981. &lt;P&gt; But what does deep learning actually do?  It is a computer
  1982. science data classification technique. It&apos;s used to take input data and
  1983. classify it: give it a thing and it will figure out what the thing is.
  1984. But it classifies things in a way that&apos;s different and more useful than
  1985. traditional computer science methods for classification, such as computer
  1986. programming, or data storage and retrieval (databases). As such, it can be
  1987. used to do a lot more than computers previously had been able to do.  &lt;P&gt;
  1988. To see this, consider traditional computer science methods: for example,
  1989. computer programming. This approach requires a person to write code that
  1990. explicitly considers different cases. For example, imagine that you want to
  1991. classify two-dimensional figures.  You want to consider whether they are
  1992. regular polygons. You could write a computer program that defines for itself
  1993. what a regular polygon is, and checks each characteristic of an input shape
  1994. to see whether or not it matches the definition of a regular polygon. Such a
  1995. program, when given a square, will notice that it is a polygon, it has four
  1996. sides, and that those sides are equal in length. Since the programmer
  1997. put into the program a detailed definition of what a regular polygon is,
  1998. and since the program checks each feature explicitly, it can tell whether
  1999. or not a shape is a regular polygon, even if the program has never seen
  2000. that particular shape before.
  2001.  
  2002. &lt;P&gt;But what about exceptional cases? Is a circle a regular polygon? It is,
  2003. after all, the limit of an N-gon as N goes to infinity. This is an &quot;edge
  2004. case&quot; and programs need to consider those explicitly. A programmer had to
  2005. anticipate this case and write it into the program. Moreover, if you wanted to
  2006. consider some other type of shape, a programmer would have to rewrite
  2007. the code accordingly. There&apos;s no going from a bunch of examples to working
  2008. code without a programmer to write it.  Programming is certainly a useful
  2009. technique, but it has its limits. Wouldn&apos;t it be nice to be able to
  2010. learn from a bunch of examples, without a person having to write all that
  2011. code?
  2012.  
  2013. &lt;P&gt; One way to do that would be data storage and retrieval, for example,
  2014. a database.  Consider the shape classifier problem again. You might put
  2015. in a bunch of shapes into a database, indicating whether the shape is
  2016. a regular polygon or not. Once the database is populated, classifying a
  2017. shape simply becomes looking it up. The database will say whether or not
  2018. it is a regular polygon.
  2019.  
  2020. &lt;P&gt; But what if it&apos;s not there? A database has the advantage of being
  2021. able to learn from examples. But it has a big disadvantage: if it hasn&apos;t
  2022. seen an example before, and is asked about it, it has no idea what the
  2023. right answer is. So while data storage and retrieval is a very useful
  2024. computing technique, and it is the backbone of most of our modern
  2025. information systems, it has its limits. Wouldn&apos;t it be nice if
  2026. a classifier system could provide a useful answer for input data
  2027. that it&apos;s never seen before, without a programmer to tell it how?
  2028.  
  2029. &lt;P&gt; Deep learning does exactly this. Like data storage and retrieval, it
  2030. learns from examples, through training. Very roughly, a neural network,
  2031. when trained, is given some input data, and is told what output
  2032. data it should produce when it sees that data in future. These input and
  2033. output constraints propagate forward and backwards through the network,
  2034. and are used to modify internal values such that when the network next
  2035. sees input like that, it will produce the matching output.
  2036.  
  2037. &lt;P&gt; The key advantage of this technique is that if it sees data that
  2038. is similar to, but not the same as data it has been trained on, it will
  2039. produce output similar to the trained output. This is very important,
  2040. because like programming, it can work on input it has never seen, but like
  2041. databases, it can learn from examples and need not be coded by a programmer
  2042. anticipating all the details in advance. For our shape example, if trained
  2043. with many examples of regular polygons, the neural network will be able to
  2044. figure out whether or not a given input is a regular polygon, and perhaps
  2045. even more interestingly, it will be able to note that a circle is very
  2046. like a regular polygon, even if it had never been trained on a circle.
  2047.  
  2048. &lt;P&gt; Moreover, a deep learning neural network can learn from its own
  2049. results. This is called reinforcement learning. This technique involves
  2050. using a neural network to derive output data from some input data, the
  2051. results are tested to see how well they work, and the neural network is
  2052. retrained accordingly.  This way a neural network can &quot;learn from its own
  2053. mistakes&quot;, training itself iteratively to classify better. For example, a
  2054. model of a walking human, with some simple programming to teach it the laws
  2055. of physics, can, using reinforcement learning, teach itself how to walk. A
  2056. few years ago, some of the researchers in our department
  2057. did exactly that.  Another example: Google got a lot of
  2058. attention a few years ago when deep learning researchers there &lt;A
  2059. HREF=&quot;https://deepmind.com/research/case-studies/alphago-the-story-so-far&quot;&gt;
  2060. built a deep learning system that used reinforcement learning to become a
  2061. champion at the game of Go, a game very hard to computerize using traditional
  2062. techniques, and proved it by beating the reigning Go world champion. &lt;/A&gt;
  2063.  
  2064.  
  2065. &lt;P&gt; It seems clear to me at this point that deep learning is as fundamental
  2066. a computing technique as computer programming and databases in building
  2067. practical computer systems. It is enormously powerful, and is causing a
  2068. great deal of legitimate excitement. Like all computer science techniques, it
  2069. has its advantages and drawbacks, but its strengths are where other computer
  2070. science techniques have weaknesses, and so it is changing computer science
  2071. (and data science more generally) in dramatic ways. It&apos;s an interesting time
  2072. to be a computer scientist, and I can&apos;t even begin to imagine the many things
  2073. that bright and innovative people will be able to do with it in the future.
  2074. </description>
  2075.  </item>
  2076.  <item>
  2077.    <title>Existential threats from AI?</title>
  2078.    <pubDate>Mon, 02 Sep 2019 20:14:00 -0400</pubDate>
  2079.    <link>http://www.cs.toronto.edu/%25252525252525252525252525257Ejdd/blog.cgi/2019/09/02#aithreats</link>
  2080.    <category>/it</category>
  2081.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525257Ejdd/blog.cgi/it/aithreats</guid>
  2082.    <description>&lt;figure style=&quot;float:left;padding:1px&quot;&gt;
  2083. &lt;IMG src=&quot;/~jdd/blog/it/nuclear-2136244_640.jpg&quot; alt=&quot;Nuclear explosion&quot;&gt;
  2084. &lt;figcaption&gt;
  2085. Image by &lt;a href=&quot;https://pixabay.com/users/AlexAntropov86-2691829/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=2136244&quot;&gt;Alexander Antropov&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=2136244&quot;&gt;Pixabay&lt;/a&gt;
  2086. &lt;/figcaption&gt;
  2087. &lt;/figure&gt;
  2088.  
  2089. Plenty has been written about the possible threats to humanity from Artificial
  2090. Intelligence (AI). This is an old concern, a staple of science fiction since
  2091. at least the 1950s. The usual story: a machine achieves
  2092. sentience and pursues its own agenda, harmful to people. The current successes
  2093. of machine learning have revived this idea. The &lt;A HREF=&quot;https://www.bbc.com/news/technology-30290540&quot;&gt;late Stephen
  2094. Hawking warned the BBC in 2014&lt;/A&gt; that &quot;the development of full artificial
  2095. intelligence could spell the end of the human race&quot;.  He feared that &quot;it
  2096. would take off on its own, and re-design itself at an ever increasing rate.&quot;
  2097. He worries that human beings, &quot;who are limited by slow biological evolution, couldn&apos;t compete, and would be superseded.&quot;
  2098. &lt;A
  2099. HREF=&quot;https://www.theatlantic.com/magazine/archive/2018/06/henry-kissinger-ai-could-mean-the-end-of-human-history/559124/&quot;&gt;Henry
  2100. Kissinger, in a thoughtful essay in &lt;em&gt;The Atlantic&lt;/em&gt; last
  2101. year&lt;/A&gt;, worried that &quot;AI, by mastering certain competencies more
  2102. rapidly and definitively than humans, could over time diminish human
  2103. competence and the human condition itself as it turns it into data.&quot;  &lt;A
  2104. HREF=&quot;https://www.wired.com/story/elon-musk-humanity-biological-boot-loader-ai/&quot;&gt;Elon
  2105. Musk, in a debate last month with Alibaba&apos;s Jack Ma, reported by WIRED,
  2106. &lt;/A&gt; argued that &quot;there&apos;s just a smaller and smaller corner of
  2107. what of intellectual pursuits that humans are better than computers. And
  2108. that every year, it gets smaller and smaller, and soon will be far far
  2109. surpassed in every single way. Guaranteed. Or civilization will end.&quot;
  2110.  
  2111. &lt;P&gt;Are they right? Is there an existential threat to humanity from
  2112. AI? Well, yes, I think there actually is one, but not quite in the way
  2113. Musk, Kissinger, or Hawking fear. Computer have been better at humans for
  2114. a long time in many cognitive domains. Computers remember
  2115. things more accurately, process things faster, and scale better
  2116. than humans in many tasks. AI, particularly machine learning, increases
  2117. the number of skills where computers are better than humans. Given that
  2118. humanity has been spending the last couple of generations getting used to
  2119. a certain arrangement where computers are good at some things and humans
  2120. are good at others, it can be a bit disconcerting to have this upended by
  2121. computers suddenly getting good at things they weren&apos;t good at before. I
  2122. understand how this can make some people feel insecure, especially highly
  2123. accomplished people who define themselves by their skills. Kissinger,
  2124. Musk and Hawking fear a world in which computers are better at many things
  2125. than humans. But we have been living in such a world for decades. AI simply
  2126. broadens the set of skills in question.
  2127.  
  2128. &lt;P&gt;As a computer scientist, I am not particularly worried about the notion of
  2129. computers replacing people. Yes, computers are developing new useful skills,
  2130. and it will take some getting used to. But I see no imminent danger of
  2131. AI resulting in an artificial person, and even if it did, I don&apos;t think
  2132. an artificial person is an intrinsic danger to humans.  Yet I agree that
  2133. there are real existential threats to humanity posed by AI. But these are
  2134. not so much long term or philosophical, to me they&apos;re eminently practical
  2135. and immediate.
  2136.  
  2137. &lt;P&gt; The first threat is the same sort of threat as posed by nuclear
  2138. physics: AI can be used to create weapons that can cause harm to people on a massive scale.  Unlike nuclear bombs, AI weapons do not do their harm through
  2139. sheer energy discharge. Rather, machine learning, coupled with advances in
  2140. miniaturization and mass production, can be used to create horrific smart
  2141. weapons that learn, swarms of lethal adaptive drones that seek out and
  2142. destroy people relentlessly. A deep commitment to social responsibility,
  2143. plus a healthy respect for the implications of such weapons, will be needed
  2144. to offset this danger.
  2145.  
  2146. &lt;P&gt; The second threat, perhaps even more serious, comes not from AI itself
  2147. but from the perceptions it creates. AI&apos;s successes are transforming
  2148. human work: because of machine learning, more and more jobs,
  2149. even white-collar ones requiring substantial training, can be replaced by
  2150. computers. It&apos;s unclear yet to what extent jobs eliminated by AI will be offset
  2151. by new jobs created by AI, but if AI results in a widespread perception
  2152. that most human workers are no longer needed, this perception may itself
  2153. become an existential threat to humanity. The increasingly obvious fact
  2154. of anthropogenic climate change has already fueled the idea that humanity
  2155. itself can be viewed as an existential threat to the planet. If AI makes
  2156. it possible for some to think that they can have the benefits of society
  2157. without keeping many people around to do the work, I worry we may see
  2158. serious consideration of ways to reduce the human population to much
  2159. smaller numbers. This to me is a dangerous and deeply troubling idea,
  2160. and I believe a genuine appreciation for the intrinsic value of all human
  2161. beings, not just those who are useful at the moment, will be needed to
  2162. forestall it. Moreover, a good argument from future utility can also be
  2163. made: we cannot accurately predict which humans will be the great inventors
  2164. and major contributors of the future, the very people we need to address
  2165. anthropogenic climate change and many other challenges. If we value all
  2166. people, and build a social environment in which everyone can flourish,
  2167. many innovators of the future will emerge, even from unexpected quarters.
  2168.  
  2169. &lt;P&gt;Threats notwithstanding, I don&apos;t think AI or machine learning
  2170. can go back into Pandora&apos;s box, and as a computer scientist who has been
  2171. providing computing support for machine learning since long before it
  2172. became popular, I would not want it to. AI is a powerful tool, and like
  2173. all powerful tools, it can be used for many good things. Let us build a
  2174. world together in which it is used for good, not harm.
  2175. </description>
  2176.  </item>
  2177.  <item>
  2178.    <title>Why we thought for a while Pluto was a planet, but it never was.</title>
  2179.    <pubDate>Mon, 26 Aug 2019 06:51:00 -0400</pubDate>
  2180.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525257Ejdd/blog.cgi/2019/08/26#pluto</link>
  2181.    <category>/misc</category>
  2182.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525257Ejdd/blog.cgi/misc/pluto</guid>
  2183.    <description>&lt;IMG src=&quot;/~jdd/blog/misc/pluto-1315109_640.jpg&quot; alt=&quot;Pluto&quot; style=&quot;float:left;padding:10px&quot;&gt;
  2184.  
  2185. &lt;P&gt; More than a decade after Pluto&apos;s demotion from the rank of planet, some
  2186. still do not accept it. I can sympathize. Like many of us, I grew up
  2187. memorizing in school the nine planets of the Solar system, the last of
  2188. which was Pluto: icy, distant and mysterious. I remember as a child poring
  2189. over a diagram of the solar system, marvelling at the concentric elipses of
  2190. the planetary orbits, and wondering why Pluto&apos;s orbit was so odd. For odd
  2191. it was: all the other planets orbited the sun in more or less concentric
  2192. elipses, but Pluto was eccentric: its orbit was at an unusual angle, and
  2193. it even briefly came closer to the sun than Neptune.  None of the other
  2194. plants had orbits like this: why Pluto? But I didn&apos;t question that it was a
  2195. planet. It had been recognized as a planet since Clyde Tombaugh discovered
  2196. it before my parents were born. For me, Pluto was weird, but it was still
  2197. &quot;planet&quot;, the astronomical equivalent of a sort of odd uncle who behaved
  2198. strangely and kept to himself, but still family.
  2199.  
  2200. &lt;P&gt; But the idea of Pluto as a planet started to become problematic in
  2201. the early 1990s. In 1992, Jewitt and Luu discovered another object beyond
  2202. Neptune: Albion, much smaller than Pluto, and also with an odd orbit. Because
  2203. it was a small object, it was pretty clearly not a planet, so Pluto&apos;s
  2204. status was not yet in question, but it was only the first of many. By 2000,
  2205. more than seventy such objects had been discovered. Most of these were very
  2206. small, but some were not so small. And the discoveries continued. In 2003,
  2207. with the discovery of the Eris, a trans-Neptunian body more
  2208. massive than Pluto itself, the problem became acute. No longer was Pluto
  2209. the odd uncle of the planets: now there were on the order of 100 odd
  2210. uncles and aunts, and at least one of them, Eris, aptly named after the greek
  2211. goddess of discord, had a better claim to planethood than Pluto itself.
  2212. Something had to be done.  This bunch of odd objects, odd in the same way as
  2213. Pluto, were either all planets, or they were none of them planets. There
  2214. was no reasonable distinction that could be made that would keep Pluto
  2215. a planet but deny planethood to Eris and many of her siblings. To do so
  2216. would be arbitrary: we would be saying that Pluto was a planet simply because
  2217. we discovered it first and it took us a long time to discover the others.
  2218. What to do?
  2219.  
  2220. &lt;P&gt; Happily, there was a precedent: this sort of thing had come up
  2221. before. In 1801, Giuseppe Piazza discovered Ceres,
  2222. a body orbiting between Mars and Jupiter. This was a big deal.  Only twenty
  2223. years before, a new planet had been discovered for the first time in
  2224. recorded history: Uranus, found by accident by William Herschel in 1781. Now,
  2225. twenty years later, Piazza had found a second.  And this one was not out beyond
  2226. Saturn, it was nearer than Jupiter. But Piazza&apos;s share of the limelight
  2227. was soon to lessen. his planet had a rival: a year later, Heinrich Wilhelm
  2228. Olbers discovered Pallas, another body between Jupiter and Mars. Two
  2229. years later, in 1804, Karl Harding discovered another: Juno. Not to be
  2230. outdone, Olbers in 1807 discovered yet another, Vesta.  By the middle of
  2231. the 19th century, fifteen bodies orbiting between Mars and Jupiter
  2232. were known, and while none of them were anywhere as large as Ceres, one
  2233. of them, Vesta, had nearly a third of Ceres&apos; mass. Were there really many
  2234. small planets between Mars and Jupiter, or were these something else?
  2235. When in 1846 the planet Neptune was discovered beyond Uranus, it became
  2236. clear that some decision about these bodies between Mars and Jupiter needed
  2237. to be made. A consensus emerged: Ceres and other such objects were not
  2238. planets. They were called &quot;asteroids&quot;, a name coined in 1802 by William
  2239. Herschel. It was a good call: there are now well over 100,000 known
  2240. asteroids, far too many for schoolchildren to memorize.
  2241.  
  2242. &lt;P&gt; With Pluto, a similar situation was now occurring. While we weren&apos;t
  2243. yet at 100,000 Pluto-like bodies, we knew about quite a few more than
  2244. fifteen. And Pluto, unlike Ceres, wasn&apos;t even the most massive: Eris was,
  2245. and quite possibly, bigger ones would be found. There was no denying the
  2246. facts. Pluto, like Ceres, could not be a planet. It must be something else.
  2247.  
  2248. &lt;P&gt; Of course this was quite controversial. People had been calling Pluto
  2249. a planet for the better part of a century. Generations of schoolchildren
  2250. had memorized it as part of the list of planets. But the choice was clear:
  2251. either the schoolchildren would have to start memorizing longer lists,
  2252. much much longer ones, or Pluto would have to be demoted. Well, not
  2253. demoted, exactly, but newly recognized for what it really was all
  2254. along: something different. In the sumer of 2006, the &lt;A
  2255. HREF=&quot;http://www.iau.org&quot;&gt;International Astronomical Union
  2256. (IAU)&lt;/A&gt; declared that Pluto isn&apos;t a planet, it is a &lt;A
  2257. HREF=&quot;https://www.iau.org/static/resolutions/Resolution_GA26-5-6.pdf&quot;&gt;&lt;EM&gt;dwarf
  2258. planet&lt;/EM&gt;&lt;/A&gt;. While this designation is a little confusing (if a
  2259. dwarf planet isn&apos;t a planet, why is it called a dwarf &lt;EM&gt;planet&lt;/EM&gt;?),
  2260. one thing was now clear: Pluto is not the same sort of thing as Mercury,
  2261. Venus, Earth, Mars, Jupiter, Saturn, Uranus and Neptune; it, and Eris,
  2262. and probably a couple of other larger trans-Neptunian bodies discovered
  2263. since the 1990s, are something different.  But guess what: Ceres, too,
  2264. fits IAU&apos;s definition of dwarf planet, the only asteroid that does. Two
  2265. centuries after its discovery, Ceres, first-born of the non-planets
  2266. and largest of the asteroids, was deemed a dwarf planet, and Piazza, its
  2267. discoverer, though not the second person in recorded history to discover
  2268. a new planet, was recognized as the very first to discover a dwarf one.
  2269. </description>
  2270.  </item>
  2271.  <item>
  2272.    <title>Ross Anderson&apos;s &lt;EM&gt;Security Engineering&lt;/EM&gt;</title>
  2273.    <pubDate>Fri, 19 Jul 2019 16:13:00 -0400</pubDate>
  2274.    <link>http://www.cs.toronto.edu/%252525252525252525252525252525257Ejdd/blog.cgi/2019/07/19#anderson</link>
  2275.    <category>/it</category>
  2276.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525252525257Ejdd/blog.cgi/it/anderson</guid>
  2277.    <description>&lt;A HREF=&quot;https://www.cl.cam.ac.uk/~rja14/book.html&quot;&gt;
  2278. &lt;IMG src=&quot;https://www.cl.cam.ac.uk/~rja14/Papers/book2coversmall.jpg&quot; alt=&quot;Security Engineering - Second Edition&quot; style=&quot;float:left;padding:10px&quot;&gt;&lt;/A&gt;
  2279.  
  2280. &lt;P&gt;Until recently, I had not read Ross Anderson&apos;s &lt;EM&gt;Security
  2281. Engineering&lt;/EM&gt;, despite hearing good things about it. I&apos;m not sure why:
  2282. I think I was put off a bit by the title. I had a vague and confused
  2283. impression that a book about &quot;Security Engineering&quot; would be yet another
  2284. how-to book about making computers secure. I should have known better. In
  2285. this case, I was wrong, very much so, and much to my detriment. I should
  2286. have read this book long ago.
  2287.  
  2288. &lt;P&gt; Why had I not read it?  I have no excuse. The book has
  2289. been out for a while: it is in its second edition, which came
  2290. out in 2008 (Anderson is writing a third edition, expected next
  2291. year). So I certainly had the opportunity.  Moreover, since 2012,
  2292. the book has been free for the reading (and downloading) from &lt;A
  2293. HREF=&quot;https://www.cl.cam.ac.uk/~rja14/book.html&quot;&gt;his website&lt;/A&gt;. So I
  2294. certainly had the means.  I just didn&apos;t, until a few weeks ago, when I
  2295. stumbled across it again. I read a little from the website, then a little
  2296. more. Before long, I was well and thoroughly hooked.
  2297.  
  2298. &lt;P&gt;&lt;EM&gt;Security Engineering&lt;/EM&gt; is a classic, comprehensive book about
  2299. information security: eminently readable, clear and thorough, it covers
  2300. information security in pretty much every aspect one might encounter it,
  2301. from the usual (cryptography, access controls, protocols, biometrics) to the
  2302. not quite so day-to-day (nuclear weapons launch protocols, counterfeiting,
  2303. even spying by analyzing the RF emissions from computers). Each chapter
  2304. is a clear elucidation of a particular aspect of information security,
  2305. focusing on the essential issues. Each chapter provides enough detail to
  2306. understand the essential elements, yet not too much detail as to overwhelm
  2307. the reader. His writing is a classic illustration of the difference between
  2308. an expert and a master. An expert knows a great deal about a topic and
  2309. provides an abundance of information. A master knows the key elements,
  2310. those things that are most important, on which everything else hangs, and
  2311. focuses exactly on these. This book is mastery, in clear, understandable
  2312. and engaging language. It has become my favourite book in information
  2313. security already, and I haven&apos;t yet finished it.
  2314.  
  2315. &lt;P&gt;I look forward to the third edition sometime next year. I can&apos;t wait.
  2316. </description>
  2317.  </item>
  2318.  <item>
  2319.    <title>Externality and Information Security</title>
  2320.    <pubDate>Mon, 04 Mar 2019 12:04:00 -0500</pubDate>
  2321.    <link>http://www.cs.toronto.edu/%25252525252525252525252525252525257Ejdd/blog.cgi/2019/03/04#externality</link>
  2322.    <category>/it</category>
  2323.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525252525257Ejdd/blog.cgi/it/externality</guid>
  2324.    <description>
  2325. It was a hot midsummer weekend, and I was traveling back to Toronto
  2326. with friends. We were on the expressway (the name here in Ontario for
  2327. the sort of road that Americans call freeways and Brits call motorways).
  2328. Traffic was very slow: a classic traffic jam. After about thirty minutes,
  2329. we reached the cause of the problem. It was not a collision. Nor was it
  2330. highway construction. Instead, by the side of the roadway, a minivan was
  2331. parked, back gate open, and a family was having a picnic on the nearby
  2332. grass. I don&apos;t know if they realized they were causing a traffic jam, but
  2333. they were. People had slowed to look, which caused traffic behind to slow
  2334. too, and because of the traffic volume, this led to a traffic jam over a
  2335. considerable distance.
  2336.  
  2337. &lt;P&gt; I don&apos;t know why the family having the picnic had chosen that spot for
  2338. it, and I don&apos;t know whether they realized the problem they were causing.
  2339. But their picnic went on, unaffected by the traffic problems they were
  2340. causing. In other words, the traffic jam was &lt;em&gt;not their problem&lt;/em&gt;.
  2341. It was an &lt;em&gt;externality&lt;/em&gt;, something causing a negative effect not
  2342. felt by those who cause it.
  2343.  
  2344. &lt;P&gt; Externalities happen in life all the time. Large organizations
  2345. (companies, countries, institutions) suffer significantly when their
  2346. decision-makers make decisions that are good for themselves but not good
  2347. for the organization. Rules to make this less likely are put in place:
  2348. rules against bribery, rules concerning conflict of interest, rules imposing
  2349. due process. But rules only work to a certain extent: there are plenty of
  2350. situations where the rules are followed yet still externalities happen.
  2351. Moreover, rules come with costs, sometimes significant ones.  Rules may
  2352. be necessary, but they are not sufficient, and they need to be accompanied
  2353. by buy-in.
  2354.  
  2355. &lt;P&gt;Let&apos;s consider traffic again. Driving is governed by all sorts of rules.
  2356. Some of these rules work well: at traffic lights, go when the light is
  2357. green, stop when it is red.  Rarely broken, this rule makes traffic work
  2358. in dense situations where otherwise there would be chaos. Most of the
  2359. time, this rule is followed even in the absence of external enforcement.
  2360. When enforcement does occur, it is well regarded: hardly anyone will argue
  2361. that a person running a red light is a safety hazard and should be ticketed.
  2362. In practice, you can stand for hours beside a busy traffic signal in a
  2363. typical Ontario city, and despite the absence of police presence, not
  2364. find a single driver running a red light.
  2365.  
  2366. &lt;P&gt;Sadly, other driving rules don&apos;t work quite so well, such as speed limits
  2367. on expressways here in Ontario. These limits are often broken, with some
  2368. following them and others not. Often, on an uncongested expressway,
  2369. unless enforcement is likely (i.e. police is present) there will be some
  2370. people driving over the speed limit. Enforcement is viewed cynically:
  2371. speeding tickets are often viewed more as revenue generation than as a
  2372. safety measure. Obeying speed limits is often viewed by drivers as
  2373. an externality: &lt;em&gt;not my problem&lt;/em&gt;, unless there is a police officer
  2374. around to make it one. In practice, at any place on any uncongested
  2375. Ontario expressway, you will be hard-pressed to find a five-minute period
  2376. in which no passing driver has exceeded the speed limit.
  2377.  
  2378. &lt;P&gt;I have been thinking a lot about information security lately. In
  2379. information security, we have a situation similar in many respects to
  2380. driving. Just as driving is a matter of traveling safely, information
  2381. security is a matter of computing safely. When we compute, we may
  2382. be processing information that is sensitive, confidential, private. Harm
  2383. can occur when it is exposed.  Steps need to be taken to ensure that it
  2384. is not: persons handling information will have to handle it securely. But
  2385. do we want this process to look like speed limits? Or traffic lights? I
  2386. think the answer is clear: if we want information to actually be secure,
  2387. we want good security practice to be followed like the rules for traffic
  2388. lights are followed: broadly and consistently, without the need for the
  2389. constant threat of enforcement.
  2390.  
  2391. &lt;P&gt;In recent years, an information security profession has
  2392. arisen.  The increasing demands of the profession have made it increasingly
  2393. rare that an information security professional has spent much time actually
  2394. running a substantial IT operation.  Certifications abound, and a
  2395. multiplicity of complex and large security standards have been created,
  2396. each requiring professionals to interpret.  A great deal of money is
  2397. being spent on information security. Much of this is good and necessary:
  2398. information security needs attention, codification, dissemination, and
  2399. championship.  But the professionalization of information security comes
  2400. with big risks, too: the risk that information security will become the
  2401. responsibility only of specialists, the risk that these specialists will
  2402. come up with all-encompassing codexes of security standards to impose,
  2403. the risk that these standards will be treated as externalities by IT
  2404. practitioners, the risk that the information security profession will
  2405. respond with enforcement, and hence the risk we will find ourselves in
  2406. the expressway speed limit situation with respect to information security.
  2407.  
  2408. &lt;P&gt;The fact is, information security is an aspect of good IT practice:
  2409. if an implementation is not secure, it is &lt;em&gt;broken&lt;/em&gt;, just as much
  2410. as if it were not reliable. Security is the responsibility of all IT
  2411. practitioners: it needs to be internalized, not externalized.
  2412.  
  2413. &lt;P&gt;For this to happen, it is important that information security rules
  2414. be simple and understandable, to ensure buy-in.  Just as traffic light
  2415. rules address the obvious risk of traffic accidents, so should security
  2416. rules address clear risks in a visibly appropriate way. In most cases,
  2417. it&apos;s not so important that rules be part of a comprehensive codex that
  2418. addresses all possible areas of risk: the more complex the rule and the
  2419. more extensive the system of rules, the more likely it will all be treated
  2420. as an externality. What we really want are not rules for their own sake,
  2421. but genuinely secure IT.
  2422.  
  2423. &lt;P&gt; If we want secure IT, we need to recognize that there is another
  2424. potential externality at work. Genuine information security and the good of
  2425. the information security profession may not always align. Just as expressway
  2426. speed limits employ more police than traffic lights, an enforcement approach
  2427. will employ more information security professionals than an internalized
  2428. one. But the internalized approach is what gives us secure computing. This
  2429. is not something that can be left to the information security profession
  2430. alone. To get there, we will need collaborative effort from all of us,
  2431. particularly those with long experience running substantial IT operations.
  2432. We will all need to make a true commitment to a practical approach, one
  2433. that seeks to make computing genuinely more secure in the real world.
  2434. </description>
  2435.  </item>
  2436.  <item>
  2437.    <title>Intentionality</title>
  2438.    <pubDate>Tue, 26 Feb 2019 06:27:00 -0500</pubDate>
  2439.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525252525257Ejdd/blog.cgi/2019/02/26#intentionality</link>
  2440.    <category>/misc</category>
  2441.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525252525257Ejdd/blog.cgi/misc/intentionality</guid>
  2442.    <description>
  2443. &lt;p&gt; I spent all of 2018 intending to blog, and not doing it. Sadly, this
  2444. is an all too human situation. We intend to do things, when we can, when
  2445. time permits, but we can&apos;t; time doesn&apos;t permit. Or at least this is
  2446. one of those stories we tell ourselves. The truth is a little simpler:
  2447. throughout 2018, my intention to blog was not strong enough for me to
  2448. re-prioritize things in my day so that I would do it.
  2449.  
  2450. &lt;p&gt; I had plenty to say. I continue to have plenty to say. I had plenty
  2451. of important things to do, and that also continues to be true. Despite my
  2452. other responsibilities, I am making time now, and I will continue to
  2453. make time, every so often, to say things in this blog. I am being
  2454. &lt;em&gt;intentional&lt;/em&gt; about it.
  2455.  
  2456. &lt;p&gt; To be intentional about something means to be deliberately purposeful:
  2457. to make one&apos;s actions a directly chosen consequence of one&apos;s thoughtful
  2458. decisions. For most people, myself included, life is full of input,
  2459. distractions, demands, requests. It is easy to fill time without much
  2460. effort. But if I am not intentional, it will be filled with reaction,
  2461. not action: things that circumstances and prior commitments have chosen
  2462. for me, not things I have chosen for myself.
  2463.  
  2464. &lt;p&gt; Reaction is fine, even good and necessary. Many people, myself included,
  2465. build up throughout their lives various important responsibilities:
  2466. responsibilities to family, work, friends, communities. Responsibilities
  2467. carry with them a commitment to react to the needs of others. This is
  2468. well and good. But it is not enough, at least not for me. I realize that
  2469. to be authentic, I have to consider carefully what is important to me,
  2470. decide what to do about it, and then act on it. This is intentionality. I&apos;ve
  2471. decided to be intentional about blogging. Look for more
  2472. blog entries in the coming weeks.
  2473. </description>
  2474.  </item>
  2475.  <item>
  2476.    <title>A Way to Visualize Relative Masses of Things in the Solar System </title>
  2477.    <pubDate>Tue, 12 Dec 2017 13:07:00 -0500</pubDate>
  2478.    <link>http://www.cs.toronto.edu/%252525252525252525252525252525252525257Ejdd/blog.cgi/2017/12/12#astromass</link>
  2479.    <category>/misc</category>
  2480.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525252525252525257Ejdd/blog.cgi/misc/astromass</guid>
  2481.    <description>
  2482. Every so often we hear things in the news about the solar system: a mission
  2483. to a planet or asteroid, talk of manned missions to mars, arguments about
  2484. whether Pluto is a planet or not.  We tend to have pretty sketchy ideas of
  2485. what most bodies in the solar system are like compared to Earth.  The fact
  2486. is that they&apos;re more wildly different in size and mass than we might think.
  2487.  
  2488. &lt;P&gt;Let&apos;s look at mass.  Imagine you decide to row across San Francisco
  2489. bay in a 12-foot aluminum rowboat.  You pack a couple of suitcases, your
  2490. 15 inch Macbook Pro (can&apos;t go without connectivity) and your ipad mini,
  2491. you get in your rowboat and start rowing.  As you row, you get hungry, so
  2492. you pull out a Snickers bar.  Now imagine that the USS Nimitz, a massive
  2493. nuclear-powered aircraft carrier, passes by.  There you are, in a rowboat
  2494. with your two suitcases, your Macbook Pro, your iPad, and your Snickers bar,
  2495. alongside a huge supercarrier.
  2496.  
  2497. &lt;P&gt; Well, the mass of the sun compared to the earth is like that aircraft
  2498. carrier compared to you and your boat.  The mass of Mars is like your
  2499. two suitcases.  The mass of the moon is like your 15 inch Macbook Pro,
  2500. and the mass of Pluto is like your iPad mini.  As for the Snickers bar,
  2501. it&apos;s like Ceres, the largest of the asteroids.
  2502.  
  2503. &lt;P&gt;Now let&apos;s suppose the massive wake of the aircraft carrier tips over
  2504. your rowboat and leaves you in the water.  Along comes a rich tech founder
  2505. in his 70 foot yacht, and fishes you out.  That yacht is like Jupiter,
  2506. the largest planet.
  2507.  
  2508. &lt;P&gt;So forget any mental images you might have of planets being something
  2509. like the Sun, only a bit smaller and cooler.  The sizes of things in the
  2510. solar system are really quite different, and there is nothing, absolutely
  2511. nothing, in the solar system that is anything quite like the Sun.
  2512. </description>
  2513.  </item>
  2514.  <item>
  2515.    <title>Bitcoin, Cryptocurrency and Blockchain</title>
  2516.    <pubDate>Mon, 11 Dec 2017 14:02:00 -0500</pubDate>
  2517.    <link>http://www.cs.toronto.edu/%25252525252525252525252525252525252525257Ejdd/blog.cgi/2017/12/11#bitcoin</link>
  2518.    <category>/it</category>
  2519.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525252525252525257Ejdd/blog.cgi/it/bitcoin</guid>
  2520.    <description>
  2521. &lt;P&gt;As the price of Bitcoin goes up and up, talk increases about Bitcoin
  2522. and other cryptocurrencies, like Litecoin, Monero, ZCash, Ethereum,
  2523. and many others. Plenty is being said, and it can be a bit confusing.
  2524.  
  2525. &lt;P&gt;But there is no need to be confused.  Bitcoin and other cryptocurrencies
  2526. are basically simple.  They are not coins.  They are simply lists.
  2527. Each cryptocurrency has a master list.  The list typically contains
  2528. information about &lt;em&gt;who&lt;/em&gt; and &lt;em&gt;what&lt;/em&gt; (i.e. amounts).  The list
  2529. is designed in a clever way, using computer software, so that people
  2530. all over the world can have identical copies of the list and keep it up
  2531. to date, without someone having to be the holder of the &quot;master copy&quot;.
  2532. But it is still just a list.  
  2533.  
  2534. &lt;P&gt; The sort of list used for cryptocurrencies is called a &quot;blockchain&quot;,
  2535. and it has some special properties.  One particularly clever property is
  2536. that you can&apos;t normally just add anything you want to the list, there is a
  2537. scheme to control that.  Instead, you need to arrange with someone already
  2538. on the list to give up (some of) their place on the list to you.
  2539.  
  2540. &lt;P&gt;So when someone says they bought some Bitcoin and they&apos;re going to make
  2541. a lot of money, what they mean (whether they realize it or not) is that
  2542. they paid somebody some money to put them on a list, and they hope that
  2543. someone later will pay them even more money to get off it.  
  2544.  
  2545. &lt;P&gt;As for me, I haven&apos;t &quot;bought&quot; any. As I write this, cryptocurrency
  2546. prices are rising fast.  But I think what is happening is a kind of run-away
  2547. positive feedback loop: people are buying in because it is going up, and it
  2548. is going up because people are buying in.  Eventually it will run out of
  2549. people to buy in, and it will stop going up.  Then some people will sell,
  2550. causing the feedback loop to go the other way: people will sell because
  2551. it is going down, and it will go down because people are selling.
  2552.  
  2553. &lt;P&gt;That being said, one thing in particular about cryptocurrency is
  2554. making me grumpy about it, even though I don&apos;t &quot;own&quot; any. Recall I wrote
  2555. that you can&apos;t &lt;em&gt;normally&lt;/em&gt; make yourself a new entry on a blockchain
  2556. list, but there is a way.  You can do an enormous lot of computations on
  2557. a computer in an attempt to find new special numbers that can be used
  2558. to create new entries on the list. This process is misnamed &quot;mining&quot;,
  2559. but it&apos;s more a sort of computerized brute-force mathematical searching.
  2560. Those computations take a long time and use a lot of electricity.  Moreover,
  2561. even the ordinary transactions generated by people &quot;buying&quot; and &quot;selling&quot;
  2562. a cryptocurrency is a computational burden, since there
  2563. are so many copies of the list around the world.  Each list
  2564. is very big: Bitcoin&apos;s is more than 100GB, and every copy
  2565. need to be updated.  This uses electricity too. In fact, &lt;A
  2566. HREF=&quot;https://digiconomist.net/bitcoin-energy-consumption&quot;&gt;digiconomist.net
  2567. estimates that Bitcoin computations alone presently use up enough
  2568. electricity to power more than three million US households.&lt;/A&gt;
  2569. Furthermore, the &quot;mining&quot; computers use GPUs that are
  2570. really good for graphics and machine learning, but because cryptocurrency
  2571. &quot;miners&quot; are buying them all up, those GPUs are getting harder to
  2572. find for a good price. Personally, I am not happy with the
  2573. challenges I am having in finding enough GPU resources for our computer
  2574. scientists, who are hungry for GPUs for machine learning.  While high
  2575. demand for GPUs is maybe good for GPU manufacturers (for example, &lt;A
  2576. HREF=&quot;http://fortune.com/2017/09/03/bitcoin-graphics-card-makers-stocks/&quot;&gt;according
  2577. to fortune.com, Nvidia made U$150M in one quarter in 2017 selling GPUs
  2578. to cryptocurrency &quot;miners&quot;&lt;/A&gt;), surely all those GPUs, and all that
  2579. electricity, can be used for something more useful than cryptocurrency.
  2580. </description>
  2581.  </item>
  2582.  <item>
  2583.    <title>A closer look at topuniversities.com&apos;s 2017 rankings for Computer Science.</title>
  2584.    <pubDate>Thu, 09 Mar 2017 12:58:00 -0500</pubDate>
  2585.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525252525252525257Ejdd/blog.cgi/2017/03/09#rankings2017</link>
  2586.    <category>/misc</category>
  2587.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525252525252525257Ejdd/blog.cgi/misc/rankings2017</guid>
  2588.    <description>
  2589. &lt;p&gt;&lt;A HREF=&quot;http://topuniversities.com&quot;&gt;The QS World University Rankings&lt;/A&gt;
  2590. for 2017 are out, including the subject rankings.  For the subject &quot;Computer
  2591. Science &amp; Information Systems&quot;, the University of Toronto does very well,
  2592. placing tenth.
  2593.  
  2594. &lt;P&gt;A closer look at the top ten shows some expected leaders (MIT, Stanford,
  2595. CMU, UC Berkeley) but some less expected ones, such as Oxford and Cambridge.
  2596. These are superb Universities with good Computer Science programs, but
  2597. are their CS programs really among the ten best in the world?
  2598.  
  2599. &lt;P&gt;A closer look at how the score is computed sheds some light on this
  2600. question.  The Overall Score is a combination of Academic Reputation,
  2601. Citations per Paper, Employer Reputation, and H-index Citations.
  2602. Academic Reputation and Employer Reputation are, in essence, the opinions of
  2603. professors and employers respectively. While (hopefully) they are reasonably
  2604. well founded opinions, this is a subjective, not an objective, metric.
  2605. On the other hand, Citations per Paper and H-index Citations are objective.
  2606. So I looked at Citations per Paper and H-index Citations for the top forty
  2607. schools on the 2017 QS Computer Science &amp; Information Systems ranking.
  2608.  
  2609. &lt;P&gt;By Citations per Paper, top five of those forty are:
  2610. &lt;OL&gt;
  2611. &lt;LI&gt;Princeton
  2612. &lt;LI&gt;Stanford
  2613. &lt;LI&gt;UT Austin
  2614. &lt;LI&gt;Washington
  2615. &lt;LI&gt;UC Berkeley
  2616. &lt;/OL&gt;
  2617.  
  2618. &lt;P&gt;No MIT?  This seems off.  So lets look at the top five by H-Index Citations:
  2619.  
  2620. &lt;OL&gt;
  2621. &lt;LI&gt;Stanford
  2622. &lt;LI&gt;MIT
  2623. &lt;LI&gt;UC Berkeley
  2624. &lt;LI&gt;UI Urbana-Champaign
  2625. &lt;LI&gt;UT Austin
  2626. &lt;/OL&gt;
  2627.  
  2628. &lt;P&gt;That looks more reasonable.  So let&apos;s look at the top twenty by H-Index Citations:
  2629. &lt;OL&gt;
  2630. &lt;LI&gt;Stanford
  2631. &lt;LI&gt;MIT
  2632. &lt;LI&gt;UC Berkeley
  2633. &lt;LI&gt;UI Urbana-Champaign
  2634. &lt;LI&gt;UT Austin
  2635. &lt;LI&gt;Georgia IT
  2636. &lt;LI&gt;CMU
  2637. &lt;LI&gt;Tsinghua
  2638. &lt;LI&gt;Nanyang
  2639. &lt;LI&gt;ETH Zurich
  2640. &lt;LI&gt;Washington
  2641. &lt;LI&gt;Princeton
  2642. &lt;LI&gt;UBC
  2643. &lt;LI&gt;Toronto
  2644. &lt;LI&gt;Waterloo
  2645. &lt;LI&gt;NU Singapore
  2646. &lt;LI&gt;UC London
  2647. &lt;LI&gt;Cornell
  2648. &lt;LI&gt;UCLA
  2649. &lt;LI&gt;CU Hong Kong
  2650. &lt;/OL&gt;
  2651.  
  2652. &lt;P&gt;That&apos;s a list that makes more sense to me.  While it puts my department
  2653. 14th instead of 10th, I think I have more confidence in the objectivity
  2654. of this ordering than I do in the QS Overall Score ordering.
  2655. </description>
  2656.  </item>
  2657.  <item>
  2658.    <title>Program Source Code Should be Readable by Human Beings By Definition</title>
  2659.    <pubDate>Thu, 02 Feb 2017 13:35:00 -0500</pubDate>
  2660.    <link>http://www.cs.toronto.edu/%252525252525252525252525252525252525252525257Ejdd/blog.cgi/2017/02/02#python3</link>
  2661.    <category>/it</category>
  2662.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/python3</guid>
  2663.    <description>
  2664. Version 3 of the &lt;A HREF=&quot;https://www.python.org&quot;&gt;Python programming
  2665. language&lt;/A&gt; made a seemingly innocuous change to the Python programming
  2666. language: &lt;A HREF=&quot;https://www.python.org/dev/peps/pep-0008/#id18&quot;&gt;no
  2667. longer could tabs and spaces be mixed for indentation: either tabs must be
  2668. used exclusively, or spaces.&lt;/A&gt; Hence the following is not a valid Python
  2669. 3 program:
  2670. &lt;pre&gt;
  2671. def hello():
  2672. print(&quot;Hello&quot;)
  2673.        print(&quot;World&quot;)
  2674. hello()
  2675. &lt;/pre&gt;
  2676. If I run it, here&apos;s what I get:
  2677. &lt;pre&gt;
  2678. % python3 testme.py
  2679.  File &quot;testme.py&quot;, line 3
  2680.    print(&quot;World&quot;)
  2681.                 ^
  2682. TabError: inconsistent use of tabs and spaces in indentation
  2683. &lt;/pre&gt;
  2684. However, the following is a valid Python 3 program:
  2685. &lt;pre&gt;
  2686. def hello():
  2687.        print(&quot;Hello&quot;)
  2688.        print(&quot;World&quot;)
  2689. hello()
  2690. &lt;/pre&gt;
  2691. &lt;pre&gt;
  2692. % python3 testme.py
  2693. Hello
  2694. World
  2695. &lt;/pre&gt;
  2696. and so is the following:
  2697. &lt;pre&gt;
  2698. def hello():
  2699. print(&quot;Hello&quot;)
  2700. print(&quot;World&quot;)
  2701. hello()
  2702. &lt;/pre&gt;
  2703. &lt;pre&gt;
  2704. % python3 testme.py
  2705. Hello
  2706. World
  2707. &lt;/pre&gt;
  2708.  
  2709. Confused yet?  
  2710.  
  2711. &lt;p&gt;As you can, or perhaps more to the point, &lt;em&gt;can&apos;t&lt;/em&gt; see, the problem
  2712. here is that the first program uses a tab to indent the first
  2713. &lt;em&gt;print&lt;/em&gt; statement, and spaces to indent the second &lt;em&gt;print&lt;/em&gt;
  2714. statement.  The second program uses spaces to indent both, and the third
  2715. program uses tabs to indent both.  But because tabs and spaces are both
  2716. visually represented as whitespace, it is difficult or impossible to
  2717. visually distinguish between a correct and an incorrect python3 program
  2718. through inspecting the source code.  This breaks the basic definition of
  2719. source code: &lt;em&gt;human-readable computer instructions&lt;/em&gt;.
  2720.  
  2721. &lt;P&gt;No doubt the Python 3 designers have good intentions: to help
  2722. python programmers be consistent about indentation.  But to me, it
  2723. seems unreasonable to have a programming language where syntactically
  2724. or semantically important distinctions are not clearly visible in the
  2725. source code.
  2726. </description>
  2727.  </item>
  2728.  <item>
  2729.    <title>Slow Windows Update on Windows 7 again?  Install two Windows Update patches first.</title>
  2730.    <pubDate>Wed, 23 Nov 2016 09:48:00 -0500</pubDate>
  2731.    <link>http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525257Ejdd/blog.cgi/2016/11/23#wupdate2</link>
  2732.    <category>/it</category>
  2733.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/wupdate2</guid>
  2734.    <description>
  2735. Back in May, &lt;A
  2736. HREF=&quot;http://www.cs.toronto.edu/~jdd/blog.cgi/2016/05/09#wupdate&quot;&gt;
  2737. I wrote&lt;/A&gt; about Windows Update for Windows 7 taking many hours or even
  2738. days; the fix then was to install two patches manually first.
  2739.  
  2740. &lt;P&gt; The problem has returned. Even if you install the two patches I mentioned
  2741. in May, you may experience very slow updates on Windows 7.
  2742.  
  2743. &lt;P&gt;Happily, again there&apos;s a workaround: grab two patches,
  2744. different than before, and manually install them. Get &lt;A
  2745. HREF=&quot;https://support.microsoft.com/en-ca/kb/3172605&quot;&gt;
  2746. KB3172605&lt;/A&gt; and its prerequisite &lt;A
  2747. HREF=&quot;https://support.microsoft.com/en-ca/kb/3020369&quot;&gt;KB3020369&lt;/A&gt; from the
  2748. &lt;A HREF=&quot;https://www.microsoft.com/download&quot;&gt;Microsoft Download Center&lt;/A&gt;,
  2749. and install them manually in numeric order, before running Windows update.
  2750. If making a fresh Windows 7 installation, simply install Windows 7 SP1,
  2751. followed by KB3020369, then KB3172605, and only then run windows update.
  2752. These two patches seem to address the slowness issues: after they were
  2753. installed on some of our systems here, Windows Update ran in a reasonable
  2754. amount of time.
  2755. </description>
  2756.  </item>
  2757.  <item>
  2758.    <title>Remembering Kelly Gotlieb</title>
  2759.    <pubDate>Wed, 26 Oct 2016 10:41:00 -0400</pubDate>
  2760.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2016/10/26#gotlieb</link>
  2761.    <category>/misc</category>
  2762.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525257Ejdd/blog.cgi/misc/gotlieb</guid>
  2763.    <description>
  2764. &lt;P&gt;On October 16th, 2016, &lt;A
  2765. HREF=&quot;http://web.cs.toronto.edu/news/current/In_Memoriam__The__father_of_computing_in_Canada__Calvin_C__Gotlieb.htm&quot;&gt;Kelly
  2766. Gotlieb&lt;/A&gt;, founder of the &lt;A HREF=&quot;http://www.cs.toronto.edu&quot;&gt;Department
  2767. of Computer Science&lt;/A&gt; at the &lt;A HREF=&quot;http://www.utoronto.ca&quot;&gt;University
  2768. of Toronto&lt;/A&gt;, passed away in his 96th year.  I had the privilege
  2769. of knowing him.  Kelly was a terrific person: brilliant, kind, and
  2770. humble. He was always willing to make time for people.  He was a great
  2771. thinker: his insights, particularly in the area of computing and society,
  2772. were highly influential.  I never fully realized how influential
  2773. he was until we, here at the department of Computer Science,
  2774. created a &lt;A HREF=&quot;http://socialissues.cs.toronto.edu&quot;&gt;blog,
  2775. http://socialissues.cs.toronto.edu&lt;/A&gt;,
  2776. in honour of the 40th anniversary of &lt;em&gt;
  2777. &lt;A HREF=&quot;https://books.google.ca/books?id=xLTiBQAAQBAJ&quot;&gt;Social
  2778. Issues in Computing&lt;/A&gt;&lt;/em&gt;, the seminal textbook he and &lt;A
  2779. HREF=&quot;http://www.cs.toronto.edu/~bor&quot;&gt;Allan Borodin&lt;/A&gt; wrote in 1973
  2780. in the area of computers and society.  I served as editor of the blog,
  2781. and solicited contributions from the top thinkers in the field. So many
  2782. of them responded, explaining to me how influential his ideas had been to
  2783. them, and the blog was filled with insightful articles building in various
  2784. ways upon the foundation that he and Allan had laid so many years before. I
  2785. interviewed Kelly for the blog, and he was terrific: even in his nineties,
  2786. he was full of insights. His mind active and enthusiastic, he was making
  2787. cogent observations on the latest technologies, ranging from self-driving
  2788. cars to automated medical diagnosis and treatment.
  2789.  
  2790. &lt;P&gt;To me, Kelly epitomized the truth about effective teaching that is all too
  2791. often missed: teaching is not just about information, teaching is about
  2792. inspiration.  Kelly was a truly inspiring teacher and thinker.  He was
  2793. completely authentic in everything he did, he was full of enthusiasm,
  2794. and that enthusiasm was infectious.  Conversations with Kelly so often
  2795. left me energized and inspired, thinking along new directions of thought
  2796. that something he said had triggered, or leaping past obstacles that
  2797. had previously seems insurmountable.  That is true teaching. Information
  2798. without inspiration is simply fodder for forgetfulness, but teaching that
  2799. inspires leads to new insights, integration of ideas, genuine understanding,
  2800. and a better, clearer and sharper window on the world. Kelly inspired so
  2801. many people for so many years.  We are truly blessed that he was among us.
  2802. He will be remembered.
  2803. </description>
  2804.  </item>
  2805.  <item>
  2806.    <title>The Price of Google</title>
  2807.    <pubDate>Sun, 16 Oct 2016 18:02:00 -0400</pubDate>
  2808.    <link>http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2016/10/16#google</link>
  2809.    <category>/it</category>
  2810.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/google</guid>
  2811.    <description>
  2812. I am a Canadian still living in the city in which I was
  2813. born. I love living in Canada, but life in Canada has its price.
  2814. &lt;A HREF=&quot;http://thecanadianencyclopedia.ca/en/article/al-purdy/&quot;&gt;Al
  2815. Purdy&lt;/A&gt;, the late 20th century Canadian poet, once wrote about Canada
  2816. as a country where everyone knows, but nobody talks about, the fact that
  2817. you can die from simply being outside.  It is true, of course: almost
  2818. everywhere in Canada, the winter is cold enough that a sufficient number
  2819. of hours outside without protection can lead to death by exposure. But
  2820. this basic fact is designed into pretty much everything in Canadian life,
  2821. it is simply accepted as a given by well over thirty million Canadians,
  2822. and we cope: we wear the right winter clothes, we heat and insulate our
  2823. buildings in winter, we equip our cars with the right tires, and life goes
  2824. on. Despite the Canadian winter, Canada is a great place to live.
  2825.  
  2826. &lt;P&gt;Google offers a lot of very good free web services: it is &quot;a great
  2827. place to live&quot; on the Internet, and their services are used by hundreds
  2828. of milliions of people all over the world.  While Google seems about as
  2829. far removed from a Canadian winter as you can imagine, there&apos;s something
  2830. in their Terms of Service that people seem to rarely talk about, something
  2831. that might have a bit of a chilling effect on one&apos;s initial ardor.
  2832.  
  2833. &lt;P&gt;Google, to its credit, has a very clear and easy-to-read &lt;A
  2834. HREF=&quot;https://www.google.com/intl/en/policies/terms/&quot;&gt;Terms of
  2835. Service document&lt;/A&gt;. Here&apos;s an excerpt from the version of April 14, 2014,
  2836. which is the most current version at the time I write this.
  2837.  
  2838. &lt;blockquote&gt;
  2839. When you upload, submit, store, send or receive content to or through our
  2840. Services, you give Google (and those we work with) a worldwide license to
  2841. use, host, store, reproduce, modify, create derivative works (such as those
  2842. resulting from translations, adaptations or other changes we make so that
  2843. your content works better with our Services), communicate, publish, publicly
  2844. perform, publicly display and distribute such content. The rights you grant
  2845. in this license are for the limited purpose of operating, promoting, and
  2846. improving our Services, and to develop new ones. This license continues
  2847. even if you stop using our Services (for example, for a business listing
  2848. you have added to Google Maps).
  2849. &lt;/blockquote&gt;
  2850.  
  2851. Let me pull out for closer examination the most important bits.
  2852. For readability, I&apos;ve omitted elipses.
  2853.  
  2854. &lt;blockquote&gt;
  2855. When you submit content to our Services, you give Google (and those we work
  2856. with) a worldwide license to use such content for the purpose of our Services.  
  2857. This continues even if you stop using our Services.
  2858. &lt;/blockquote&gt;
  2859.  
  2860. &lt;P&gt;As you can see, this is pretty broad.  You are granting Google and their
  2861. partners the right to use your content for Google&apos;s Services (present and
  2862. future) anywhere in the world, forever.  While it does say that it must be
  2863. used for the purpose of their Services, it doesn&apos;t limit itself to
  2864. existing Services and it doesn&apos;t constrain what a &quot;Service&quot; might be.  Since
  2865. developing and offering Services, broadly understood, pretty much covers
  2866. the gamut of what Google does as a company, the answer is &lt;em&gt;Yes&lt;/em&gt;:
  2867. by submitting content to their services, you are granting Google and their
  2868. partners the right to use your content anywhere in the world, forever,
  2869. for a broadly unconstrained set of purposes.
  2870.  
  2871. &lt;P&gt;So does this mean nobody should use Google?  Does the Canadian winter
  2872. mean that nobody should live in Canada? After all, as Al Purdy writes,
  2873. in Canada you can die from simply being outside.
  2874.  
  2875. &lt;P&gt;Well, no, of course not.  While Google has the right to do broadly
  2876. unconstrained things with our content that we submit to them, their self
  2877. -interest is typically aligned with our&apos;s: they want us to entrust our
  2878. content to them, because they use it to earn money to operate. Therefore,
  2879. to persuade us to keep submitting content to them, they will work hard
  2880. to protect and secure the content they already have, in ways they think
  2881. we consider important.  For this reason, I think it&apos;s not unreasonable to
  2882. trust Google with some of my content: I believe they are likely to protect
  2883. it in sensible ways.  Other content I choose not to submit to Google.
  2884. Just as I am prepared for a Canadian winter, knowing it is the price
  2885. I pay to live in Canada, I continue to use some Google services, knowing that
  2886. they will keep and use my content.  Many Google services are very good and
  2887. well worth using, much of my content is not very sensitive, and I trust
  2888. Google enough to share content with them.
  2889.  
  2890. &lt;p&gt;I do wonder, however, how many Google users really understand the rights
  2891. they are granting to Google.  Canada has been around for centuries: the
  2892. Canadian winter is no secret.  But the implications of Google&apos;s broad
  2893. right to use our content are not quite so obvious.  It&apos;s not really so clear
  2894. how Google is using the content or might use it in the future, and even if
  2895. we trust Google, can we trust all those who might put pressure on Google?
  2896. Quite frankly, we really don&apos;t know yet how Google&apos;s massive repository of
  2897. our collective content can be used.  We can envision wonderful outcomes:
  2898. historians a century or two hence coming to insightful conclusions about
  2899. early twenty-first century society, for example, but we can also envision
  2900. outcomes not quite so sanguine: for example, a twenty-first century
  2901. version of Orwell&apos;s &lt;em&gt;1984&lt;/em&gt;, a dystopian world of &quot;thought-crimes&quot;
  2902. and &quot;doublespeak&quot; where content is is scanned for dissent from a prevailing
  2903. ideology.  A certain degree of caution may be warranted: in the case of
  2904. Google, unlike Canada, we may not have yet seen how severe winter can be.
  2905. A certain degree of caution is warranted.  Yes, use Google, but use it
  2906. knowing what you are doing.
  2907.  
  2908. &lt;P&gt;One last thing to be said: I focus on Google here, but
  2909. the same issues hold for Facebook, Twitter, Yahoo and other purveyors
  2910. of free services over the Internet.  Read their Terms of Service to
  2911. learn what rights you are granting by your use of their services,
  2912. and decide on the basis of that knowledge how to use their services,
  2913. and even whether you use their services at all.  After all, &lt;A
  2914. HREF=&quot;https://en.wikipedia.org/wiki/Snowbird_(person)&quot;&gt;even Canadians
  2915. sometimes choose to spend winter in Florida, Mexico, or Arizona&lt;/A&gt;.
  2916. </description>
  2917.  </item>
  2918.  <item>
  2919.    <title>The Sun-Managers Mailing list: a Knowledge Sharing Success Story</title>
  2920.    <pubDate>Mon, 16 May 2016 20:29:00 -0400</pubDate>
  2921.    <link>http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2016/05/16#sunmanagers</link>
  2922.    <category>/it</category>
  2923.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/sunmanagers</guid>
  2924.    <description>
  2925. &lt;A HREF=&quot;http://sunmanagers.cs.toronto.edu&quot;&gt;Sun-Managers&lt;/A&gt; was an email
  2926. mailing list for system administrators of computers made by &lt;A
  2927. HREF=&quot;http://en.wikipedia.org/wiki/Sun_Microsystems&quot;&gt;Sun Microsystems,
  2928. Inc&lt;/A&gt;.  The list operated from mid-1989 to the fall of 2014, and I was
  2929. privileged to be part of it for almost all of its history.
  2930.  
  2931. Sun-Managers was founded in May of 1989 by &lt;A
  2932. HREF=&quot;https://www.linkedin.com/in/william-lefebvre-936b76&quot;&gt;William (Bill)
  2933. LeFebvre&lt;/A&gt;, at &lt;A HREF=&quot;http://www.northwestern.edu&quot;&gt;Northwestern
  2934. University&lt;/A&gt;.  At the time, Bill ran &lt;A
  2935. HREF=&quot;http://ftp.lanet.lv/ftp/sun-info/sunflash/1990/Aug/20.10.sunspots&quot;&gt;Sun-Spots&lt;/A&gt;,
  2936. a digest-format mailing list for system administrators
  2937. of Sun systems, but the digest format made it difficult
  2938. for people to ask questions and get a timely response. He &lt;A
  2939. HREF=&quot;https://groups.google.com/forum/#!msg/comp.sys.sun/H9z8YwunvB4/xJFp8qOLTFUJ&quot;&gt;created
  2940. Sun-Managers&lt;/A&gt;, an unmoderated mailing list intended for
  2941. short-turnaround time questions.  This was an immediate success:
  2942. so much so that by the fall of 1989, the sheer number of messages on
  2943. the list were swamping mailboxes.  In Nov 1989, Bill instituted a &lt;A
  2944. HREF=&quot;http://sunmanagers.cs.toronto.edu/policy.txt&quot;&gt;simple policy&lt;/A&gt;:
  2945. if someone asks a question on the list, other list members were expected
  2946. to reply by email directly to the person asking the question, not to the
  2947. list. The person asking the question, in turn, was expected to summarize
  2948. the answers received, and send the summary to the list.
  2949.  
  2950. &lt;P&gt; I joined the list about this time: I had started a new job at
  2951. the &lt;A HREF=&quot;http://www.utoronto.ca&quot;&gt;University of Toronto&lt;/A&gt;&apos;s &lt;A
  2952. HREF=&quot;http://www.cs.toronto.edu&quot;&gt;Computer Science department&lt;/A&gt;, a role that
  2953. included the administration of a number of Sun workstations and servers.
  2954. I was looking for resources to help me with my Sun system administration
  2955. tasks, and this list was an excellent one.  Because of this summary policy, the
  2956. list volume was manageable enough that I could keep up, yet the turnaround
  2957. time on questions was short.  I mostly &quot;lurked&quot; at first, reading but not
  2958. replying. I felt too inexpert to answer many questions, and too shy to ask.
  2959. However, I learned a great deal from what I read. Moreover, the summaries
  2960. were archived, and &lt;A HREF=&quot;http://sunmanagers.cs.toronto.edu/year.html&quot;&gt;
  2961. this archive&lt;/A&gt; became a resource in itself, a knowledge-base of practical
  2962. information about administering Sun systems.
  2963.  
  2964. &lt;p&gt; The list grew very rapidly: 343 summaries in 1990, and over 1000 in
  2965. 1991. In August of that year, it was noted that certain questions were being
  2966. asked often, and rather than waste effort answering the same question several
  2967. times, a &lt;A HREF=&quot;http://sunmanagers.cs.toronto.edu/faq.html&quot;&gt;&quot;Frequently
  2968. Asked Questions&quot; (FAQ) file&lt;/A&gt; was instituted. The &lt;A
  2969. HREF=&quot;http://sunmanagers.cs.toronto.edu/1991/0484.html&quot;&gt;first version
  2970. &lt;/A&gt; was created by a list member from &lt;A HREF=&quot;http://www.bu.edu&quot;&gt;Boston
  2971. University&lt;/A&gt;, and quickly grow to dozens of answers.
  2972.  
  2973. &lt;p&gt; By November of 1992, the list had grown to thousands of members,
  2974. and the workload of managing the list, editing the FAQ and coaching
  2975. list members on how to follow the &lt;A HREF=&quot;http://sunmanagers.cs.toronto.edu/policy.txt&quot;&gt;list policy&lt;/A&gt;
  2976. had become significant. Many list members were not individuals,
  2977. but &quot;mail exploders&quot;: email addresses that themselves were mailing
  2978. lists going to multiple individuals at a given site.  This made
  2979. handling list membership issues more complex.  Bill LeFebvre decided
  2980. to hand the list over to others. Two list members stepped up: &lt;A
  2981. HREF=&quot;http://www.linkedin.com/in/gene-rackow-87b7b25&quot;&gt;Gene Rackow&lt;/A&gt; from
  2982. &lt;A HREF=&quot;http://www.anl.gov&quot;&gt;Argonne National Laboratory&lt;/A&gt; to run the
  2983. list software, and me, to handle the FAQ and policy work. By this time,
  2984. I had benefitted from the list for a while, and I felt it was time to
  2985. &quot;give back&quot;.  At the time, I wasn&apos;t in a position to actually
  2986. run the list: I&apos;d just taken on a new role as system manager of
  2987. the &lt;A HREF=&quot;http://www.utoronto.ca&quot;&gt;University of Toronto &lt;/A&gt;&lt;A
  2988. HREF=&quot;http://www.cs.toronto.edu&quot;&gt;Computer Science Department&lt;/A&gt;&apos;s &lt;A
  2989. HREF=&quot;http://www.cdf.toronto.edu&quot;&gt;teaching laboratories&lt;/A&gt;, and had my
  2990. hands full, but I could certainly help with content.  I was really glad to
  2991. work together with Gene, a seasoned system administrator, on this rapidly
  2992. growing list, which we moved to a system at &lt;A
  2993. HREF=&quot;http://www.anl.gov&quot;&gt;Argonne National Labs&lt;/A&gt;, where Gene worked.
  2994.  
  2995. &lt;p&gt; The list continued to grow through the 1990s.  During this time,
  2996. Sun Microsystems was quietly supportive, helping Gene with hardware (a
  2997. &lt;A HREF=&quot;http://en.wikipedia.org/wiki/SPARCstation_1&quot;&gt;Sparcstation 1&lt;/A&gt;)
  2998. as the list grew.  By 1996, over two thousand summaries a year were being
  2999. produced, peaking at 2243 in 2002.
  3000.  
  3001. In May of 1998, Gene Rackow handed over list management to Rob
  3002. Montjoy from the University of Cincinnati, who in turn handed over
  3003. list management to &lt;A HREF=&quot;https://www.linkedin.com/in/mrbill&quot;&gt;Bill
  3004. Bradford&lt;/A&gt; in November of 2000. The list was moved from Argonne
  3005. National Labs to a system in Austin run by Bill.  I continued to manage
  3006. the &lt;A HREF=&quot;http://sunmanagers.cs.toronto.edu/policy.txt&quot;&gt;list
  3007. policy&lt;/A&gt; and edit list information files, such as &lt;A
  3008. HREF=&quot;http://sunmanagers.cs.toronto.edu/before-posting.txt&quot;&gt;a
  3009. &quot;think before posting&quot; reminder&lt;/A&gt; and the FAQ which had grown to
  3010. 79 questions by December 2000. This had become a bit too large, and
  3011. so &lt;A HREF=&quot;http://sunmanagers.cs.toronto.edu/faq-2000.txt&quot;&gt;19
  3012. questions deemed less frequently asked&lt;/A&gt; were
  3013. trimmed.  A further trim was made in 2005, reducing a &lt;A
  3014. HREF=&quot;http://sunmanagers.cs.toronto.edu/faq-2005.txt&quot;&gt;65-question FAQ&lt;/A&gt;
  3015. to &lt;A HREF=&quot;http://sunmanagers.cs.toronto.edu/faq.txt&quot;&gt;one under 60&lt;/A&gt;.
  3016.  
  3017. &lt;p&gt; By 2002, the list had reached over five thousand members and the
  3018. workload of running the list software and managing the list
  3019. subscriptions had become too much for one person.  Dan Astoorian, my
  3020. colleage at the &lt;A HREF=&quot;http://www.utoronto.ca&quot;&gt;University of Toronto&lt;/A&gt;,
  3021. stepped in to help, and he was sorely needed.  Moreover, the list server
  3022. hardware was feeling the strain: by mid-2001, list members were being
  3023. asked to contribute used equipment to upgrade the server.  This was
  3024. resolved in April 2003, when the list was migrated to a machine at the
  3025. &lt;A HREF=&quot;http://www.utoronto.ca&quot;&gt;University of Toronto&lt;/A&gt; that had been
  3026. donated to the University by Sun Microsystems.
  3027.  
  3028. &lt;p&gt; But times were changing.  Linux was growing rapidly and Sun&apos;s business
  3029. was being affected. The web provided more resources for people seeking
  3030. help administering their systems, and fewer were relying on mailing lists.
  3031. The list fell below 2000 summaries per year in 2003, under 1200 in 2004, and
  3032. dropped below 1000 in 2005. By 2008, summaries per year had fallen to about
  3033. 300, fewer than in any full-year period previously. Sun Microsystems ran into
  3034. significant difficulties during the economic downturn that year,
  3035. and was sold to
  3036. &lt;A HREF=&quot;http://www.oracle.com&quot;&gt; Oracle&lt;/A&gt; the following year.  As for the list, in 2009,
  3037. there were just over 200 summaries, declining to less than 100 in 2011.
  3038. More disturbingly, the ratio of summaries to questions was steadily declining,
  3039. from over 24% in 2001 to less than 16% by 2010: for some reason, list
  3040. members were becoming less diligent in summarizing responses back to the
  3041. list.  Summaries and list traffic in general continued to decline rapidly:
  3042. there were just over 50 summaries in 2012, and less than a dozen in 2013.
  3043. In 2014, there were only three by October, when a hardware failure provided
  3044. a good excuse to retire the list.
  3045.  
  3046. &lt;p&gt; The Sun-Managers mailing list, over its twenty-five year lifetime,
  3047. provided help to many thousands of system administrators, producing over
  3048. 29000 summaries, &lt;A HREF=&quot;http://sunmanagers.cs.toronto.edu/year.html&quot;&gt;an
  3049. archive of which continues to be available&lt;/A&gt;.  Special thanks is due to
  3050. the superb people I was privileged to work together with on the list over
  3051. the years: William LeFebvre, Gene Rackow, Rob Montjoy, Bill Bradford, and
  3052. Dan Astoorian.  Gratitude, also, is due to the thousands of list members
  3053. who so freely shared their knowledge and expertise with others.
  3054.  
  3055. &lt;P&gt; The list summary archive, and an account of the list&apos;s
  3056. history (on which this blog entry is based) is available at &lt;A
  3057. HREF=&quot;http://sunmanagers.cs.toronto.edu&quot;&gt;http://sunmanagers.cs.toronto.edu&lt;/A&gt;.
  3058. The list&apos;s official web page, &lt;A
  3059. HREF=&quot;http://www.sunmanagers.org&quot;&gt;http://www.sunmanagers.org&lt;/A&gt;, continues to be
  3060. maintained by Bill Bradford.
  3061. </description>
  3062.  </item>
  3063.  <item>
  3064.    <title>Slow Windows Update on Windows 7?  Install two Windows Update patches first.</title>
  3065.    <pubDate>Mon, 09 May 2016 10:54:00 -0400</pubDate>
  3066.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2016/05/09#wupdate</link>
  3067.    <category>/it</category>
  3068.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/wupdate</guid>
  3069.    <description>
  3070. Recently, I noticed Windows Update taking many hours or even days on
  3071. Windows 7, especially for new installs/reinstalls.  Task manager shows
  3072. svchost.exe exhibiting large memory usage (suggestive of a memory leak)
  3073. and/or sustained 100% CPU.
  3074.  
  3075. &lt;P&gt;Happily, there&apos;s a workaround: grab a couple of patches
  3076. to Windows Update itself, and manually install them. Get &lt;A
  3077. HREF=&quot;https://support.microsoft.com/kb/3050265&quot;&gt;KB3050265&lt;A&gt; and &lt;A
  3078. HREF=&quot;https://support.microsoft.com/kb/3102810&quot;&gt;KB3102810&lt;/A&gt; from
  3079. the &lt;A HREF=&quot;https://www.microsoft.com/download&quot;&gt;Microsoft Download Center&lt;/A&gt;,
  3080. and install them manually in that order, before running Windows update.
  3081. These two patches seem to address the issues: after they were installed
  3082. on some of our systems here, Windows Update ran in a reasonable amount of
  3083. time (an hour or two perhaps on slow systems when many updates are needed,
  3084. but not days).
  3085. </description>
  3086.  </item>
  3087.  <item>
  3088.    <title>Apple vs FBI: it is about setting a precedent.</title>
  3089.    <pubDate>Fri, 04 Mar 2016 10:25:00 -0500</pubDate>
  3090.    <link>http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2016/03/04#applefbi</link>
  3091.    <category>/it</category>
  3092.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/applefbi</guid>
  3093.    <description>
  3094. There seems to be lots of confusion about Apple&apos;s current dispute with
  3095. the FBI, despite &lt;A HREF=&quot;http://www.apple.com/customer-letter&quot;&gt;Apple&apos;s
  3096. message to their customers of Feb 16, 2016&lt;/A&gt;, where they tried to
  3097. explain the issue.  Here&apos;s the issue in a nutshell.
  3098.  
  3099. &lt;p&gt; The FBI has an Apple iPhone that was the work-phone of a now-dead
  3100. terrorist.  The FBI wants to read what is on that phone.  But the
  3101. phone is encrypted, and runs a secure version of iOS.  The FBI wants Apple
  3102. to make an insecure version of iOS to run on that phone, so that the FBI
  3103. can break into the phone and read the contents.  Apple has, so far, refused.
  3104.  
  3105. &lt;p&gt; This issue will no doubt be addressed in the US courts and legislatures.
  3106. What is at stake is the precedent it sets.  The essential question is this:
  3107. to what extent should law enforcement be able to compel others to assist
  3108. them with an investigation?  Should software developers be expected to
  3109. make insecure versions of their software, so that law enforcement can
  3110. &quot;break in&quot;?  It will be very interesting to see how this plays out.
  3111. </description>
  3112.  </item>
  3113.  <item>
  3114.    <title>Apple&apos;s new Macbook laptop: like a tablet?</title>
  3115.    <pubDate>Fri, 13 Mar 2015 11:08:00 -0400</pubDate>
  3116.    <link>http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2015/03/13#macbook</link>
  3117.    <category>/it</category>
  3118.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/macbook</guid>
  3119.    <description>
  3120. &lt;P&gt;I rarely write about Apple&apos;s products because they have no shortage of
  3121. press already: Apple has superb marketing, and many of their products are
  3122. remarkable in one way or another, often for excellent design and engineering.
  3123. Their new super-thin &lt;A HREF=&quot;http://www.apple.com/macbook/&quot;&gt;Macbook
  3124. laptop&lt;/A&gt; is no exception: it&apos;s very thin and light, has a superb
  3125. high-resolution screen, a carefully redesigned trackpad and keyboard,
  3126. and is very power-efficient.  New to this machine is the fact that it
  3127. has only a single USB-C port for power, data, and video (it also has a
  3128. headphone port for audio).  Most laptops have many more ports than this.
  3129. A USB port used for both power and data, and a headphone port, but nothing
  3130. else, is more typical of a tablet, not a laptop.
  3131.  
  3132. Indeed, some of the press seems to have
  3133. really latched onto this &quot;tablet&quot; comparison.  &lt;A
  3134. HREF=&quot;http://www.foxnews.com/tech/2015/03/11/apple-has-designed-macbook-thats-almost-tablet/&quot;&gt;Brooke
  3135. Crothers of Foxnews/Tech claims that the MacBook is &quot;almost a tablet&quot;
  3136. and states that the MacBook &quot;is an iPad with a keyboard&quot; &lt;/A&gt;while  &lt;A
  3137. HREF=&quot;http://www.slate.com/blogs/future_tense/2015/03/09/apple_releases_2_lb_12_retina_macbook.html&quot;&lt;/A&gt;Lily
  3138. Hay Newman of Slate claims that &quot;you should think of the new macbook as
  3139. a tablet&quot;&lt;/A&gt;.  So how true is this?  Is the new MacBook like a tablet?
  3140.  
  3141. &lt;P&gt;Well, no, it&apos;s not.  The MacBook&apos;s screen is not
  3142. touch-capable, and is not capable of being used like a tablet screen.
  3143. The keyboard and touchpad is an integral part of the machine: it is
  3144. not optional or detachable.  It runs a desktop/laptop operating system
  3145. (MacOSX), not a tablet operating system such as iOS.  The device is not
  3146. a tablet, it is not &quot;almost a tablet&quot;, it is not even like
  3147. a tablet.  It&apos;s a small, light, power-efficient laptop.  If it
  3148. must be compared to something, perhaps it can be compared to a &lt;A
  3149. HREF=&quot;http://www.cs.toronto.edu/~jdd/blog.cgi/2009/02/09#netbooks&quot;&gt;netbook&lt;/A&gt;,
  3150. though it has a much better keyboard, touchpad and screen, and is much
  3151. more expensive.
  3152.  
  3153. &lt;P&gt;Then what about the single I/O port?  That&apos;s simply the consequence of
  3154. the new &lt;A HREF=&quot;http://www.usb.org/developers/docs/usb_31_030215.zip&quot;&gt;USB
  3155. 3.1 specification&lt;/A&gt;, which finally allows a USB connection to deliver
  3156. enough power to power a laptop, and defines the USB-C connector, which in
  3157. addition to USB data lines, provides &quot;alternate mode&quot; data lines that can
  3158. be used for display protocols like DisplayPort.  This makes it possible for
  3159. Apple to build &lt;A
  3160. HREF=&quot;http://store.apple.com/us/product/MJ1K2AM/A/usb-c-digital-av-multiport-adapter&quot;&gt;multiport
  3161. adapters for the Macbook&lt;/A&gt; that provide video (e.g. HDMI), data (USB-A)
  3162. and charging ports, making it unnecessary to provide all those ports
  3163. separately in the laptop itself.
  3164.  
  3165. &lt;P&gt;So does this make the Macbook &quot;like a tablet&quot;?  While it is true that
  3166. tablets have been using single connectors for power and data for a long
  3167. time, this doesn&apos;t make the Macbook tablet-like.  It&apos;s not the presence of
  3168. a single shared power/data connector that makes something like a tablet,
  3169. it&apos;s the interactive screen.  Yes, a horse has four legs and is often sat
  3170. upon, but a horse is not anything like a chair.
  3171.  
  3172. &lt;P&gt; So will I be getting one of the new Macbooks? Probably not: like a fine
  3173. thoroughbred, the new Macbook is lovely but rather too expensive for me.
  3174. The need to buy the multiport adapter separately makes the already high
  3175. cost of acquisition even higher.  The high price doesn&apos;t stop me from
  3176. admiring the design and engineering of this new laptop, but it does keep
  3177. me from buying one.
  3178. </description>
  3179.  </item>
  3180.  <item>
  3181.    <title>What&apos;s wrong with Blackberry? (and some ideas about how to fix it)</title>
  3182.    <pubDate>Sat, 05 Oct 2013 17:03:00 -0400</pubDate>
  3183.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2013/10/05#blackberry</link>
  3184.    <category>/it</category>
  3185.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/blackberry</guid>
  3186.    <description>Blackberry is in the news a fair bit these days, and the news seems to be all bad.  
  3187. &lt;A
  3188. HREF=&quot;http://press.blackberry.com/financial/2013/blackberry-reports-second-quarter-fiscal-2014-results.html&quot;&gt;As
  3189. the firm reports close to a billion dollars in quarterly losses&lt;/a&gt;, a &lt;A
  3190. HREF=&quot;http://www.computerworld.com/s/article/9242767/Update_Gartner_tells_IT_shops_that_it_s_game_over_for_BlackBerry&quot;&gt;Gartner
  3191. analyst recommends that enterprise customers find alternatives to Blackberry
  3192. over the next six months&lt;/A&gt;.  What&apos;s the problem?
  3193. &lt;p&gt;
  3194. Basically, fewer and fewer people want to buy Blackberry phones.  The problem
  3195. isn&apos;t so much that Blackberries don&apos;t do what they&apos;re supposed to, it&apos;s
  3196. that people now perceive iPhones and various Android phones as much better
  3197. choices, and are buying those instead.  Why?  The reason is that an iPhone or
  3198. an Android phone isn&apos;t the same sort of phone as a traditional Blackberry. An
  3199. iPhone or an Android phone is a true smartphone, i.e. an &quot;app&quot; phone, a
  3200. platform that runs a whole &quot;ecosystem&quot; of third party software.  A traditional
  3201. Blackberry is a &quot;messaging&quot; phone, a device that specializes in effective
  3202. messaging, such as email.  Yes, it can run applications too, but that&apos;s
  3203. not its primary function, and it shows.  
  3204. &lt;p&gt;
  3205. To illustrate, consider email.  Sending email requires the ability to
  3206. type quickly.  A physical keyboard works best for this, one that stretches
  3207. across the short side of the phone.  The screen, located above the keyboard,
  3208. then becomes roughly square: it can&apos;t be very wide, because the phone will
  3209. then become too wide to hold easily or to fit in one&apos;s pocket, and it can&apos;t
  3210. be very tall or the phone will become too long.   A square screen is fine
  3211. for messaging, but for other things that a smartphone might like to do,
  3212. such as displaying video, one wants a screen that is significantly wider
  3213. than it is tall.  A smartphone handles this by having a rectangular screen:
  3214. when doing messaging, one holds the phone vertical: the bottom half of the
  3215. screen then turns into a keyboard, and the top half turns into a roughly
  3216. square messaging display.  When watching media, such as videos, the phone
  3217. is held horizontal, allowing a screen that is wider than it is tall.
  3218. Hence the smartphone is useful in a broader set of ways: it is not just a
  3219. messaging device.  Smartphones have become good enough at messaging that
  3220. many people do not feel they need a dedicated messaging device.  Once the
  3221. smartphone is the only device that people feel they need to carry, there&apos;s
  3222. much less demand for a messaging phone.
  3223.  
  3224. &lt;p&gt;
  3225. Blackberry realized the problem, and tried to create a smartphone of its
  3226. own. For instance, in 2008, it released the Blackberry Storm.  But it
  3227. became clear that Blackberry&apos;s phone OS was not as well suited for
  3228. general smartphone use as iOS and Android. The Storm was not a commercial
  3229. success because it did not work as well as competing phones.  In response,
  3230. in 2010 Blackberry bought a company called QNX that had a powerful OS,
  3231. and started building devices to use it: first the Playbook, released in
  3232. spring 2011, and then the Z10 phone in early 2013, followed a few months
  3233. later by the Q10 and other phone models.
  3234.  
  3235. &lt;p&gt;
  3236. The new Blackberry OS works better than the old in delivering
  3237. smartphone apps, but it was not very mature in 2011, and was
  3238. available only on a tablet (the Blackberry Playbook). Unfortunately,
  3239. the Playbook did not sell particularly well because Blackberry &lt;A
  3240. HREF=&quot;http://press.blackberry.com/press/2010/pressrelease-4577.html&quot;&gt;
  3241. badly misrepresented it, calling it the &quot;best professional-grade table in
  3242. the industry&quot;&lt;/A&gt; though it lacked many features of the market-leading iPad,
  3243. including key messaging features such as a standalone email client.
  3244. While it could have been a market success if it were marketed as a
  3245. Blackberry phone accessory, a role it could effectively play, at release
  3246. it was clearly not a true general-purpose tablet like the iPad.  So it
  3247. accumulated few apps, while Apple&apos;s iOS and Google&apos;s Android accumulated
  3248. many.  Blackberry realized this fairly quickly, and released an Android
  3249. application emulation environment for their OS in early 2012, which allowed
  3250. many Android apps to be easily moved over to the new OS.  But few Android
  3251. developers bothered to make Blackberry versions of their Android apps,
  3252. given the relatively few Playbooks sold.
  3253.  
  3254. &lt;p&gt; In the meanwhile, Blackberry did itself no favours by making it clear
  3255. that there was no future for its existing phones, while failing to deliver
  3256. a phone running its new OS for more than a year.  This merely encouraged
  3257. Blackberry users and app developers alike to switch to another platform.
  3258. When the Z10 phone finally came out in 2013, the bulk of its apps were
  3259. those that had been written for or ported to the Playbook, a far less rich set
  3260. of applications than any Android or iOS phone.  And while the Z10 is a decent
  3261. phone that comes with some very nice messaging features, Blackberry did
  3262. not do an effective job of touting the unique features of the Z10 that
  3263. iPhones and Android phones do not have.  Moreover, the price was set high
  3264. (about the same as an iPhone or high end Android phone) and Blackberry
  3265. produced a huge number, expecting to sell a great many.  Some sold, but
  3266. many didn&apos;t, and Blackberry&apos;s recent $1B loss was due primarily to writing
  3267. down the value of unsold Z10s.
  3268.  
  3269. &lt;p&gt;
  3270. Blackberry sits today in a difficult position.  No, it is not about to go
  3271. out of business: the company is debt-free and has a couple of billion dollars
  3272. in the bank.  But its smartphone is not selling.  What should it do now?
  3273.  
  3274. &lt;p&gt;
  3275. Blackberry&apos;s best chance at this point to make its smartphone platform
  3276. viable is to take its large inventories of written-down Z10 phones and
  3277. sell them cheaply, using a renewed marketing campaign that focuses on the
  3278. unique features of the phone&apos;s software.  The Z10 hardware is really no
  3279. different than the various Android and iPhone models out there: if the
  3280. phone is to sell, it has to be on the basis of what makes it unique, and
  3281. that&apos;s the Blackberry OS software.  For instance, Blackberry should show
  3282. everyone the clever virtual keyboard that supports fast one-handed typing,
  3283. the unique messaging hub, and the &quot;Blackberry Balance&quot; software that lets
  3284. you separate work items from personal items on the phone.  Blackberry needs
  3285. to hire the best marketing people in the world to help get the message out.
  3286. This is a &quot;make or break&quot; situation for the platform.
  3287.  
  3288. &lt;p&gt;
  3289. Secondly, Blackberry should modify the OS to run Android apps natively,
  3290. without repackaging.  Android app developers are not going to repackage
  3291. their apps for Blackberry.  Blackberry needs to recognize this and make
  3292. sure that Android apps will appear automatically on Blackberry devices.
  3293. Blackberry will need to find a way to get Google Play (the Android app store)
  3294. ported to the platform.  It is too late to build a separate app ecosystem
  3295. around the Blackberry OS: it has to leverage an existing ecosystem, or die.
  3296. Android is really the only viable option for Blackberry right now.
  3297.  
  3298. &lt;p&gt;
  3299. Finally, Blackberry needs to recognize that a niche market for dedicated
  3300. messaging devices exists, and continue making devices that are the best
  3301. messaging phones available, while tapping into an existing app ecosystem.
  3302. Blackberry needs to be careful not to compromise the devices&apos; effectiveness
  3303. for messaging: it should pay attention to how people use the devices in the
  3304. real world, and address quickly whatever issues they have.  If Blackberry
  3305. can&apos;t find a way of building such messaging devices using its own OS, it
  3306. should switch to Android.  Blackberry knows how to make superb messaging
  3307. phones, and it should find a way to continue to do what it does best.
  3308. </description>
  3309.  </item>
  3310.  <item>
  3311.    <title>Cloud Computing: Everything Old is New Again</title>
  3312.    <pubDate>Tue, 20 Aug 2013 22:45:00 -0400</pubDate>
  3313.    <link>http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2013/08/20#cloud</link>
  3314.    <category>/it</category>
  3315.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/cloud</guid>
  3316.    <description>
  3317. There is a great deal of hype about Cloud Computing at the moment, and
  3318. it&apos;s getting a great deal of
  3319. attention.  It&apos;s no wonder: when firms such as &lt;A
  3320. HREF=&quot;http://www.bloomberg.com/news/2013-07-21/netflix-rule-as-no-1-s-p-performer-but-.html&quot;&gt;Netflix,
  3321. with a market capitalization of over U$15B&lt;/A&gt;, use cloud computing to
  3322. deliver streaming video services to nearly forty
  3323. million customers around the world, and when the &lt;A
  3324. HREF=&quot;http://www.businessinsider.com/amazon-vs-ibm-in-a-battle-for-the-cloud-2013-7&quot;&gt;US
  3325. Central Intelligence Agency spends U$600M for cloud computing services&lt;/A&gt;,
  3326. people take notice.  But what is it all about?
  3327.  
  3328. &lt;p&gt; Cloud computing is not really a new thing, it&apos;s a variation of a very
  3329. old idea, with a new name.  In the 1960s, when computers were large and
  3330. expensive, not everyone could afford their own.  Techniques for sharing
  3331. computers were developed, and firms arose whose business was selling
  3332. time on computers to other firms.  This was most commonly described as
  3333. &quot;timesharing&quot;.  IBM released its &lt;A HREF=&quot;http://www.vm.ibm.com/history/&quot;&gt;VM
  3334. virtualization environment in 1972&lt;/A&gt;, which allowed a mainframe computer
  3335. to be divided up into virtual computers, each for a different workload.
  3336. A timesharing vendor could buy and operate an IBM computer, then rent to
  3337. their customers &quot;virtual computers&quot; that ran on that machine.  From the
  3338. customer&apos;s perspective, it was a way to obtain access to computing without
  3339. buying one&apos;s own computer.  From the vendor&apos;s perspective, it was a way of
  3340. &quot;renting out&quot; one&apos;s investment in computer infrastructure, as a viable
  3341. business.
  3342.  
  3343. &lt;p&gt;Today, cloud computing, as did timesharing in the past, involves
  3344. the renting of virtual computers to customers.  The name has changed: then,
  3345. it was called &quot;timesharing&quot;; now, &quot;cloud computing&quot;.  The type of physical
  3346. machine has changed: then, a mainframe was used to provide computing
  3347. services; now, a grid computer. The interconnection has changed: then,
  3348. leased data lines were typically used; now, the internet.  But the basic
  3349. concept is the same: a vendor rents virtual computers to customers, who
  3350. then use the virtual computers for their computing, rather than buying
  3351. their own physical computers.
  3352.  
  3353. &lt;p&gt;The advantages and disadvantages of today&apos;s cloud computing echo the
  3354. pros and cons of yesterday&apos;s timesharing.  Advantages include risk sharing,
  3355. the ability to pay for just the amount of computing needed, the option to
  3356. scale up or down quickly, the option to obtain computing resources without
  3357. having to develop and maintain expertise in operating and maintaining
  3358. those resources, and the ability to gain access to computing resources in
  3359. very large or very small quantities very quickly and easily.  Moreover,
  3360. cloud computing vendors can develop economies of scale in running physical
  3361. computers and data centres, economies that they can leverage to decrease the
  3362. cost of computing for their customers.  Disadvantages of cloud computing
  3363. include possibly higher unit costs for resources (for example, cloud
  3364. data storage and data transfer can be very expensive, especially in
  3365. large quantities), a critical dependance on the cloud computing vendor,
  3366. variable computing performance, substantial security and privacy issues,
  3367. greater legal complexity, and so on.  These tradeoffs are neither
  3368. surprising nor particularly new: in fact, many are typical of &quot;buy&quot;
  3369. vs. &quot;rent&quot; decisions in general.
  3370.  
  3371. &lt;p&gt; Then why does cloud computing seem so new?  That, I think, is an artifact
  3372. of history.  In the 1970s and early 1980s, computers were expensive and
  3373. timesharing was popular. In the 1990s and early 2000s, computers became
  3374. increasingly cheaper, and running one&apos;s own became enormously popular.
  3375. Timesharing faded away as people bought and ran their own computers.
  3376. Now the pendulum is swinging back, not driven so much by the cost
  3377. of computers themselves, but the costs of datacentres to house them.
  3378. A few years ago, &lt;A HREF=&quot;http://www.amazon.com&quot;&gt;Amazon Inc.&lt;/A&gt; saw a
  3379. business opportunity in making virtual machines available for rental:
  3380. it was building grid computers (and datacentres to house them) for its
  3381. own operations anyway; why not rent out some of those computing resources
  3382. to other firms? In so doing, Amazon developed an important new line of
  3383. business.  At the same time, a huge number of new internet firms arose,
  3384. such as Netflix, whose operations are dominantly or exclusively that of
  3385. providing various computer-related services over the internet, and it made
  3386. a great deal of sense for such firms to use Amazon&apos;s service.  After all,
  3387. when a company&apos;s operations are primarily or exclusively serving customers
  3388. on the internet, why not make use of computing resources that are already on
  3389. the internet, rather than build private datacentres (which takes time,
  3390. money and expertise)?  These new internet firms, with lines of business
  3391. that were not even possible a decade or two ago, and Amazon&apos;s service,
  3392. also only a few years old, have lent their sheen of newness to the notion
  3393. of &quot;cloud computing&quot; itself, making it appear fresh, inventive, novel.
  3394. But is it?  The name is new, yes. But in truth, the concept is almost
  3395. as old as commercial computing itself: it has merely been reinvented for
  3396. the internet.
  3397.  
  3398. &lt;p&gt; Of course, the computing field, because of its inventiveness, high rate
  3399. of change and increasing social profile, is rather at risk of falling into
  3400. trendiness, and cloud computing certainly has become a significant trend.
  3401. The danger of trendiness is that some will adopt cloud computing not on
  3402. its own merits, but solely because it seems to be the latest tech tsunami:
  3403. they want to ride the wave, not be swamped by it.  But cloud computing
  3404. is complex, with many pros and cons; it is certainly a legitimate choice,
  3405. as was timesharing before it, but it is not necessarily the best thing for
  3406. everyone.  It&apos;s easier to see this, I think, if we look beyond the name,
  3407. beyond the trend, and see that the &quot;rent or buy&quot; question for computing
  3408. has been with us for decades, and the decision between renting virtual
  3409. machines and buying physical ones has often been complex, a balance of
  3410. risks, opportunities, and resources.  For an internet firm whose customers
  3411. are exclusively on the internet, renting one&apos;s computing assets on the
  3412. internet may make a great deal of sense.  For other firms, it may not make
  3413. sense at all.  Deciding which is true for one&apos;s own firm takes wisdom and
  3414. prudence; a healthy dose of historical perspective is unlikely to hurt,
  3415. and may help cut through the hype.
  3416. </description>
  3417.  </item>
  3418.  <item>
  3419.    <title>Handling Unsolicited Commercial Email</title>
  3420.    <pubDate>Tue, 23 Apr 2013 12:56:00 -0400</pubDate>
  3421.    <link>http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2013/04/23#uce</link>
  3422.    <category>/misc</category>
  3423.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/misc/uce</guid>
  3424.    <description>
  3425. &lt;p&gt; My email address is all over the web: at the time of writing this, &lt;A
  3426. HREF=&quot;http://www.google.ca/search?q=%22jdd%40cs.toronto.edu%22&quot;&gt;a search
  3427. on google&lt;/A&gt; for my email address produces about 15,800 results.  So
  3428. anyone who wants to find my email address can do so easily.
  3429.  
  3430. Many people or companies who want to sell me something send me email
  3431. out of the blue.  I get a great deal of such unsolicited commercial
  3432. email, too much to read or pay adequate attention to.  I simply delete
  3433. them. Unfortunately, many sources of such email persist.  So for some time
  3434. now, I&apos;ve elicited the help of technology.  I process my incoming email using
  3435. &lt;A HREF=&quot;http://www.procmail.org&quot;&gt;procmail&lt;/A&gt;, a powerful piece of software
  3436. that lets me script what happens to my email.  When I receive unsolicited
  3437. commercial email, if it is from a vendor or organization I don&apos;t have a
  3438. relationship with, I will often add a procmail rule to discard, unseen, all
  3439. future email messages from that vendor.  I&apos;ve got about 400 organizations
  3440. (mostly vendors) in my discard list so far, and the list slowly grows.
  3441. Am I still getting unsolicited commercial email from these sources?  I am,
  3442. but I am not seeing it.  It&apos;s the same effect, really, as manual deletion
  3443. (i.e.  the message is deleted, unread), but it&apos;s easier for me, because
  3444. I am not interrupted.  But of course I think it would be better still if
  3445. the email were not sent at all.
  3446.  
  3447. &lt;p&gt; If you are a vendor with whom I do not have a pre-existing relationship,
  3448. and you want to send me email introducing your products, please don&apos;t.
  3449. &lt;A HREF=&quot;blog.cgi/2011/12/15#coldcalls&quot;&gt;I do not accept cold salescalls
  3450. either.&lt;/A&gt; Instead, advertise effectively on the web, so that if I am
  3451. looking for a product like yours, I can find you.  If you must contact
  3452. me directly, send me something by postal mail, where, unlike email, the
  3453. communication does not have an interruptive aspect.
  3454. </description>
  3455.  </item>
  3456.  <item>
  3457.    <title>A closer look at the University of Toronto&apos;s international ranking in Computer Science.</title>
  3458.    <pubDate>Thu, 29 Nov 2012 00:00:00 -0500</pubDate>
  3459.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2012/11/29#rankings</link>
  3460.    <category>/misc</category>
  3461.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/misc/rankings</guid>
  3462.    <description>
  3463. &lt;p&gt;International rankings of universities seem to be all the rage
  3464. these days.  The interest seems to be fed by three rankings of
  3465. particular prominence that have emerged in the past decade. These
  3466. are &lt;A HREF=&quot;http://en.sjtu.edu.cn&quot;&gt;Shanghai Jiao Tong University&lt;/A&gt;&apos;s
  3467. &lt;A HREF=&quot;http://www.shanghairanking.com/&quot;&gt;
  3468. Academic Ranking of World Universities&lt;/A&gt; (sometimes known
  3469. as AWRU, or simply as the &quot;Shanghai Ranking&quot;), &lt;A HREF=&quot;http://www.qs.com&quot;&gt;Quacquarelli
  3470. Symonds&lt;/A&gt;&apos; &lt;A HREF=&quot;http://www.topuniversities.com&quot;&gt;QS
  3471. World University Rankings&lt;/A&gt;, and the &lt;A
  3472. HREF=&quot;http://www.timeshighereducation.co.uk/world-university-rankings/&quot;&gt;Times
  3473. Higher Education World University Rankings&lt;/A&gt;.   Part of the attractiveness
  3474. of these rankings is that they can become a way of &quot;keeping score&quot;, of
  3475. seeing how one institution does in comparison to others.
  3476.  
  3477. &lt;p&gt;My employer, the &lt;A HREF=&quot;http://www.utoronto.ca&quot;&gt;University of
  3478. Toronto&lt;/A&gt;, does quite well in these rankings, particularly my department,
  3479. &lt;A HREF=&quot;http://www.cs.toronto.edu&quot;&gt;Computer Science&lt;/A&gt;.  The subject
  3480. area of Computer Science is not ranked separately in the &lt;A HREF=&quot;&quot;&gt;Times
  3481. Higher Education World University Rankings&lt;/A&gt; (it&apos;s bundled together with
  3482. Engineering), but in the other two, Toronto has consistently ranked in the
  3483. top ten in the world each year in Computer Science, with only one exception.
  3484.  
  3485. &lt;p&gt;This exception is recent, however, and worth a closer look.  In the &lt;A
  3486. HREF=&quot;http://www.topuniversities.com&quot;&gt;QS World University Rankings&lt;/A&gt;
  3487. for Computer Science and Information Systems, Toronto dropped from &lt;A
  3488. HREF=&quot;http://www.topuniversities.com/university-rankings/world-university-rankings/2011/subject-rankings/technology/computer-science-information-systems&quot;&gt;10th
  3489. in 2011&lt;/A&gt; to &lt;A
  3490. HREF=&quot;http://www.topuniversities.com/university-rankings/world-university-rankings/2012/subject-rankings/technology/computer-science-information-systems&quot;&gt;15th
  3491. in 2012&lt;/A&gt;.  This big drop immediately raises all sorts of questions: has
  3492. the quality of Toronto&apos;s Computer Science programme suddenly plummetted?
  3493. Has the quality of Computer Science programmes at other universities suddenly
  3494. soared?  Or has the QS World University Rankings changed its methodology?
  3495.  
  3496. &lt;p&gt;To answer this question, let&apos;s look at how other universities have
  3497. changed from 2011 to 2012 on this ranking.  Many (MIT, Stanford, Berkeley,
  3498. Harvard, Oxford, Cornell, and others) stayed where they were.  Others dropped
  3499. precipitously: Cambridge University dropped from 3rd to 7th, UCLA from 8th to
  3500. 12th, and Caltech plummetted from 7th to 27th.  Some other universities
  3501. went up: Carnegie Mellon University (CMU) went from 9th to 3rd, ETH Zurich
  3502. from 11th to 8th, the National University of Singapore (NUS) from 12th
  3503. to 9th, and the Hong Kong University of Science and Technology (HKUST)
  3504. soared from 26th to 13th.  Surely these curious and significant changes
  3505. reflect a methodology change?  But what?
  3506.  
  3507. &lt;p&gt;The QS university rankings website, in the Methodology section, Academic subsection, reveals &lt;A HREF=&quot;http://www.iu.qs.com/university-rankings/subject-tables/?&quot;&gt;something of interest:&lt;/A&gt;
  3508. &lt;PRE&gt;
  3509. NEW FOR 2012 - Direct Subject Responses
  3510.  
  3511. Until 2010, the survey could only infer specific opinion on
  3512. subject strength by aggregating the broad faculty area opinions
  3513. of academics from a specific discipline. From the 2011 survey
  3514. additional questions have been asked to gather specific opinion
  3515. in the respondent&apos;s own narrow field of expertise. These responses
  3516. are given a greater emphasis from 2012.
  3517. &lt;/PRE&gt;
  3518. To understand this change, it needs to be recognized that the QS rankings
  3519. rely highly on the opinions of academics.
  3520. A large number of academics around the world are
  3521. surveyed: the QS rankings website indicates that in 2012, &lt;A
  3522. HREF=&quot;http://www.iu.qs.com/university-rankings/academic-survey-responses/&quot;&gt;46079
  3523. academic responses were received, of which 7.5% addressed Computer
  3524. Science.&quot; &lt;/A&gt;  The seemingly modest change made in 2012, to weigh more
  3525. heavily the opinions of academics in a field about their own field, given
  3526. its impact on the 2012 results for Computer Science, leads one to wonder
  3527. about the regional distribution of academics in Computer Science in
  3528. comparison to academics in other disciplines.  One significant factor may
  3529. be China.
  3530.  
  3531. &lt;p&gt;&lt;A
  3532. HREF=&quot;http://fusion.grids.cn/career/attachments/china-SIGCSE2010.pdf&quot;&gt;In
  3533. 1999, courses in the fundamentals of computer science became required
  3534. in most Chinese universities, and by the end of 2007, China had nearly a
  3535. million undergraduates studying Computer Science&lt;/A&gt;.  While QS rankings
  3536. does not indicate regional distribution by discipline for the academics
  3537. whose opinions it consults, the surge in the number of Chinese computer
  3538. scientists worldwide in the past decade almost certainly must have an
  3539. effect on the regional distribution of academics in Computer Science
  3540. as compared to other disciplines.  As such, is it any surprise to see
  3541. world universities prominent in China that possess strong Computer Science
  3542. programmes (such as HKUST and NUS) climb significantly in the rankings,
  3543. and others less prominent in China plummet?  But if a world ranking of
  3544. universities is so affected by regional shifts in those whose opinion
  3545. is being solicited, how reliable is it as an objective gage of the real
  3546. quality of a given university?
  3547.  
  3548. &lt;P&gt;Perhaps a more reliable gage of quality can be found
  3549. in the Shanghai ranking, which is not opinion-based, but &lt;A
  3550. HREF=&quot;http://www.shanghairanking.com/aboutarwu.html&quot;&gt;relies on concrete
  3551. indicators and metrics&lt;/A&gt;.  On the Shanghai ranking, the University of
  3552. Toronto consistently ranks 10th in the world in Computer Science in 2010,
  3553. 2011, and 2012.  But what does this mean, concretely?
  3554.  
  3555. &lt;P&gt;To answer these questions, we need to grapple with an
  3556. important fact: in Computer Science, the US dominates.  As a nation, &lt;A
  3557. HREF=&quot;http://books.google.ca/books?id=jv1m50j4ZAMC&quot;&gt;the
  3558. US has been enormously supportive of Computer Science ever since the field
  3559. first existed, and as a result, it has become pre-eminent in computing&lt;/A&gt;.
  3560. Nine of the top ten schools in the Shanghai ranking, and twenty of the
  3561. top twenty-five, are in the US.  For the University of Toronto to
  3562. be one of the handful of universities outside the US to break into
  3563. the top twenty-five, and the only one to break into the top ten,
  3564. is a significant accomplishment.  A chart is illustrative:  &lt;p&gt;&lt;IMG
  3565. SRC=&quot;http://www.cs.toronto.edu/~jdd/blog/misc/rankings/shanghai-cs-nonus-2012.jpg&quot; width=800 height=600&gt;
  3566.  
  3567. &lt;p&gt;Of course, the University of Toronto is in Canada, so a comparison
  3568. to other schools in Canada is also illustrative.  For Computer Science,
  3569. on the Shanghai ranking, there seems to be no close Canadian rival.
  3570. In 2012, UBC comes closest, being a only a few points short of breaking
  3571. into the top 25, but all other Canadian schools rank well back:
  3572. &lt;p&gt;&lt;IMG SRC=&quot;http://www.cs.toronto.edu/~jdd/blog/misc/rankings/shanghai-cs-can-2012.jpg&quot; width=800 height=600&gt;  
  3573.  
  3574. &lt;P&gt;Even compared to other disciplines
  3575. that have Shanghai rankings (only science, social science, and related
  3576. disciplines seem to be ranked), Toronto&apos;s pre-eminence in Computer Science
  3577. in Canada is striking:
  3578. &lt;p&gt;&lt;IMG SRC=&quot;http://www.cs.toronto.edu/~jdd/blog/misc/rankings/shanghai-can-2012.jpg&quot; width=800 height=600&gt;
  3579.  
  3580. &lt;p&gt;From a score-keeping perspective, I think we can conclude that the
  3581. University of Toronto is doing very well in Computer Science with respect
  3582. to other universities in Canada, and it is one of the few non-US
  3583. schools that can keep up with the US in this field.
  3584.  
  3585. &lt;p&gt;But all this needs to be put into perspective.  After all, rankings
  3586. are not a full picture, they&apos;re aggregations of metrics
  3587. of varying value, they represent a formulaic approach to something
  3588. (university education) that cannot always be so conveniently summarized,
  3589. and they reflect methodologies chosen by the producers of the rankings,
  3590. methodologies that may not always best reflect objective quality.
  3591. Of course, if the University of Toronto were to climb to fifth, I&apos;d be
  3592. pleased, and if it were to drop to fifteenth, I&apos;d be disappointed: surely
  3593. the score-keeper in me can be allowed this much.  But in the overall scheme
  3594. of things, what matters most for Computer Science at Toronto is not our
  3595. score on a ranking system, but the objective quality of our programme,
  3596. the learning outcomes of our students, and the impact of our research, and
  3597. these things, not our score on rankings, must always remain our top priorities.
  3598. </description>
  3599.  </item>
  3600.  <item>
  3601.    <title>Intel desktop CPU price-performance: Hyperthreading not helping?</title>
  3602.    <pubDate>Wed, 22 Aug 2012 14:07:00 -0400</pubDate>
  3603.    <link>http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2012/08/22#intelcpus</link>
  3604.    <category>/it</category>
  3605.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/intelcpus</guid>
  3606.    <description>Typically, CPU prices follow performance.  Faster CPUs command higher
  3607. prices; slower CPUs are available for less.  Recent Intel desktop CPUs
  3608. continue to show this general pattern, but there appears to be more to the
  3609. story than usual.
  3610.  
  3611. &lt;P&gt; At first glance, everything seems to be what you would expect. Using current
  3612. pricing in US$ at time of writing from &lt;A HREF=&quot;http://www.newegg.com&quot;&gt;newegg.com&lt;/A&gt;, we get:
  3613. &lt;table border=1&gt;
  3614. &lt;tr&gt;&lt;th&gt;Processor&lt;/th&gt; &lt;th&gt;PassMark&lt;/th&gt; &lt;th&gt;Price&lt;/th&gt; &lt;th&gt;PassMark/$&lt;/th&gt; &lt;th&gt;Price-Performance vs G640&lt;/th&gt;&lt;/tr&gt;
  3615. &lt;tr&gt;&lt;td&gt;Pentium G640&lt;/td&gt; &lt;td&gt;2893&lt;/td&gt;  &lt;td&gt;$79&lt;/td&gt;  &lt;td&gt;36.6&lt;/td&gt;        &lt;td&gt; 100%&lt;/td&gt;&lt;/tr&gt;
  3616. &lt;tr&gt;&lt;td&gt;i3-2120&lt;/td&gt;      &lt;td&gt;4222&lt;/td&gt;  &lt;td&gt;$125&lt;/td&gt; &lt;td&gt;33.8&lt;/td&gt;        &lt;td&gt; 92.2%&lt;/td&gt;&lt;/tr&gt;
  3617. &lt;tr&gt;&lt;td&gt;i5-3570&lt;/td&gt;      &lt;td&gt;7684&lt;/td&gt;  &lt;td&gt;$215&lt;/td&gt; &lt;td&gt;35.7&lt;/td&gt;        &lt;td&gt; 97.6%&lt;/td&gt;&lt;/tr&gt;
  3618. &lt;tr&gt;&lt;td&gt;i7-3770&lt;/td&gt;      &lt;td&gt;10359&lt;/td&gt; &lt;td&gt;$310&lt;/td&gt; &lt;td&gt;33.4&lt;/td&gt;        &lt;td&gt; 91.3%&lt;/td&gt;&lt;/tr&gt;
  3619. &lt;/table&gt;
  3620. The &lt;A HREF=&quot;http://www.cpubenchmark.net/&quot;&gt;PassMark (http://www.cpubenchmark.net/&lt;/A&gt;) to
  3621. dollar ratio is pretty consistent across all these processors, roughly 35 &amp;plusmn; 2.
  3622.  
  3623. &lt;P&gt; But what happens if we look at a more real-life benchmark?  Consider
  3624. SPEC CPU 2006 Integer (CINT2006) Baseline.  For each CPU, I used the CINT2006 Baseline results from
  3625. the most recently reported Intel reference system, as reported on &lt;A HREF=&quot;http://spec.org&quot;&gt;spec.org&lt;/A&gt;.  In the case of the G640, no
  3626. Intel reference system was reported, so I used the results for
  3627. a Fujitsu Primergy TX140 S1p.
  3628. &lt;table border=1&gt;
  3629. &lt;tr&gt;&lt;th&gt;Processor&lt;/th&gt; &lt;th&gt;CINT2006 Base&lt;/th&gt; &lt;th&gt;Price&lt;/th&gt; &lt;th&gt;CINT/$&lt;/th&gt; &lt;th&gt;Price-Performance vs G640&lt;/th&gt;&lt;/tr&gt;
  3630. &lt;tr&gt;&lt;td&gt;Pentium G640&lt;/td&gt; &lt;td&gt;34.4&lt;/td&gt;  &lt;td&gt;$79&lt;/td&gt;  &lt;td&gt;0.44&lt;/td&gt;        &lt;td&gt; 100%&lt;/td&gt;&lt;/tr&gt;
  3631. &lt;tr&gt;&lt;td&gt;i3-2120&lt;/td&gt;      &lt;td&gt;36.9&lt;/td&gt;  &lt;td&gt;$125&lt;/td&gt; &lt;td&gt;0.30&lt;/td&gt;        &lt;td&gt; 67.8%&lt;/td&gt;&lt;/tr&gt;
  3632. &lt;tr&gt;&lt;td&gt;i5-3570&lt;/td&gt;      &lt;td&gt;48.5&lt;/td&gt;  &lt;td&gt;$215&lt;/td&gt; &lt;td&gt;0.23&lt;/td&gt;        &lt;td&gt; 51.8%&lt;/td&gt;&lt;/tr&gt;
  3633. &lt;tr&gt;&lt;td&gt;i7-3770&lt;/td&gt;      &lt;td&gt;50.5&lt;/td&gt;  &lt;td&gt;$310&lt;/td&gt; &lt;td&gt;0.16&lt;/td&gt;        &lt;td&gt; 37.4%&lt;/td&gt;&lt;/tr&gt;
  3634. &lt;/table&gt;
  3635. When looking at CINT2006 Baseline, we see the price-performance ratio drop off dramatically as
  3636. the processor price increases.  We would expect this from the i3 to the i5,
  3637. since SPEC cpu int is a single job benchmark and the i3 to the i5 represents
  3638. a transition from two to four cores, but it&apos;s curious to see the dropoff
  3639. in the price-performance ratio between the G640 and the i3 (both dual-core
  3640. CPUs), and the i5 and the i7 (both quad-core CPUs).  What might be going on?
  3641.  
  3642. &lt;P&gt; A look at hyperthreading may provide some answers.  Intel hyperthreading is a feature
  3643. of some Intel CPUs that allow each physical core to represent itself to
  3644. the OS as two different &quot;cores&quot;.  If those two &quot;cores&quot; simultaneously run
  3645. code that happens to use different parts of the physical core, they can
  3646. proceed in parallel.  If not, one of the &quot;cores&quot; will block.  The i3 and
  3647. i7 CPUs offer hyperthreading, the Pentium G and i5 do not.  It turns out
  3648. that the PassMark benchmark sees significant speedups when hyperthreading
  3649. is turned on.  SPEC CINT2006, and many ordinary applications, do not.
  3650.  
  3651. &lt;P&gt; What about SPEC CINT2006 Rate Baseline, then?  The SPEC CPU Rate benchmarks measure
  3652. throughput, not just single-job performance, so maybe hyperthreading helps
  3653. more here?  Let&apos;s see:
  3654. &lt;table border=1&gt;
  3655. &lt;tr&gt;&lt;th&gt;Processor&lt;/th&gt; &lt;th&gt;CINT2006 Rate Base&lt;/th&gt; &lt;th&gt;Price&lt;/th&gt; &lt;th&gt;Rate Base/$&lt;/th&gt;&lt;th&gt;Price-Performance vs G640&lt;/th&gt;&lt;/tr&gt;
  3656. &lt;tr&gt;&lt;td&gt;Pentium G640&lt;/td&gt; &lt;td&gt;61.7&lt;/td&gt;  &lt;td&gt;$79&lt;/td&gt;  &lt;td&gt;0.78&lt;/td&gt; &lt;td&gt; 100%&lt;/td&gt; &lt;/tr&gt;
  3657. &lt;tr&gt;&lt;td&gt;i3-2120&lt;/td&gt;      &lt;td&gt;78.8&lt;/td&gt;  &lt;td&gt;$125&lt;/td&gt; &lt;td&gt;0.63&lt;/td&gt; &lt;td&gt; 80.7%&lt;/td&gt; &lt;/tr&gt;
  3658. &lt;tr&gt;&lt;td&gt;i5-3570&lt;/td&gt;      &lt;td&gt;146&lt;/td&gt;  &lt;td&gt;$215&lt;/td&gt;  &lt;td&gt;0.68&lt;/td&gt; &lt;td&gt; 87.0%&lt;/td&gt; &lt;/tr&gt;
  3659. &lt;tr&gt;&lt;td&gt;i7-3770&lt;/td&gt;      &lt;td&gt;177&lt;/td&gt;  &lt;td&gt;$310&lt;/td&gt;  &lt;td&gt;0.57&lt;/td&gt; &lt;td&gt; 73.1%&lt;/td&gt; &lt;/tr&gt;
  3660. &lt;/table&gt;
  3661. If we look at the transition from two to four cores (by comparing the i3 to
  3662. the i5), we now see that the price-performance of the i5 is better than the i3:
  3663. this is no surprise, since we are now measuring throughput, and from the i3
  3664. to the i5, we go from two to four cores.  But there still is a dropoff in
  3665. price-performance between the Pentium G and the i3, and again between the
  3666. i5 and the i7.  It&apos;s not as extreme as before, but it is still significant.
  3667. This suggests that hyperthreading may help with throughput, but not as much
  3668. as the increase in price would suggest.
  3669.  
  3670. &lt;P&gt; What does this mean, then?  It suggests the increase in price from a
  3671. non-hyperthreaded to a hyperthreaded Intel desktop processor may reflect more
  3672. an increase in PassMark performance than an increase in real performance.
  3673. Hyperthreading may have a positive effect, it seems, but typically not as
  3674. much as PassMark suggests.  At present, for best real-world price-performance
  3675. in Intel desktop CPUs, I would consider models without hyperthreading.
  3676. </description>
  3677.  </item>
  3678.  <item>
  3679.    <title>How to avoid being fooled by &quot;phishing&quot; email.</title>
  3680.    <pubDate>Tue, 26 Jun 2012 16:56:00 -0400</pubDate>
  3681.    <link>http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2012/06/26#phishing</link>
  3682.    <category>/it</category>
  3683.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/phishing</guid>
  3684.    <description>
  3685. A &quot;phishing&quot; email is an email message that tries to convince you to
  3686. reveal your passwords or other personal details.   Most often, it tries
  3687. to send you to a website that looks like the real thing (e.g. your bank or
  3688. your email provider) but is really a clever duplicate of the real website
  3689. that&apos;s set up by crooks to steal your information.  Often the pretence looks
  3690. authentic.  If you fall for it and give your password or other personal
  3691. details, criminals may steal your identity, clean out your bank account,
  3692. send junk email from your email account, use your online trading account
  3693. to buy some penny stock you never heard of, send email to all the people
  3694. in your address book telling them you&apos;re stranded in a foreign country and
  3695. need them to wire money immediately, or do any number of other bad things.
  3696.  
  3697. &lt;p&gt; But there&apos;s a really easy way to avoid being fooled by phishing messages.
  3698. If you get a message that asks you to confirm or update your account details,
  3699. &lt;em&gt;never, ever go to the website using a link that is in the email message
  3700. itself.&lt;/em&gt; Remember, anyone can send you a message with any sort of
  3701. fraudulent claim, containing any number of links that pretend to go to one
  3702. place, but really go to another.  So if you feel you must check, go to the
  3703. website that you know for sure is the real thing: use your own bookmark
  3704. (or type in the URL yourself), not the link in the message.  
  3705. </description>
  3706.  </item>
  3707.  <item>
  3708.    <title>Dealing with unsolicited salescalls (cold calls).</title>
  3709.    <pubDate>Thu, 15 Dec 2011 15:14:00 -0500</pubDate>
  3710.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2011/12/15#coldcalls</link>
  3711.    <category>/misc</category>
  3712.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/misc/coldcalls</guid>
  3713.    <description>
  3714. &lt;p&gt; For many years, I&apos;ve been plagued by unsolicited salescalls. It&apos;s
  3715. not very hard to find my phone number, and various people (mostly in
  3716. the IT realm) call me up out of the blue hoping to sell me something.
  3717. The interruption is unwelcome, even if the product isn&apos;t.  
  3718.  
  3719. &lt;p&gt; For some years now, my policy is to explain to the caller that I don&apos;t
  3720. accept unsolicited salescalls, sincerely apologize, and end the call.
  3721. Occasionally, I am then asked how I am to be contacted.  I explain that I
  3722. prefer to do the contacting myself: when I have a need, I am not too shy
  3723. to contact likely vendors and make inquiries about their products.
  3724.  
  3725. &lt;p&gt; Occasionally I run into someone who is offended by my unwillingness to
  3726. take their unsolicited salescall.  I do feel more than a little sympathy
  3727. for the salesperson when this happens: I imagine they may think I objected
  3728. to something they did, or to their manner.  The fact is, I handle all
  3729. unsolicited salescalls this way.  As for whether it is intrinsicly
  3730. offensive to reject unsolicited salescalls out of hand, I don&apos;t think
  3731. it is.  Indeed, it is natural for a salesperson to want their salescall,
  3732. even if unsolicited, to be better accepted.  But it is unreasonable for
  3733. any salesperson to expect that unsolicited sales inquiries to strangers
  3734. will always be welcome.  But I do apologize, each time, and in general,
  3735. when I so quickly end telephone conversations with salespersons who call
  3736. me out of the blue.
  3737.  
  3738. &lt;p&gt;Dear reader, if you are a salesperson, and you are tempted to contact
  3739. me to sell me something, please do not call.  Instead, just advertise
  3740. generally (and if you must, send me some mail in the post).  Trust me to
  3741. find you when the need arises.  I frequently do.
  3742. </description>
  3743.  </item>
  3744.  <item>
  3745.    <title>Gigabit ethernet, and Category 5, 5e cabling.</title>
  3746.    <pubDate>Tue, 26 Jul 2011 17:15:00 -0400</pubDate>
  3747.    <link>http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2011/07/26#cat5e</link>
  3748.    <category>/it</category>
  3749.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/cat5e</guid>
  3750.    <description>
  3751. There seems to be lots of folklore that says that Category 5 (Cat5)
  3752. cabling can&apos;t run gigabit ethernet.  Contrary to widespread belief,
  3753. that&apos;s mostly false.  Here&apos;s the situation.  Cat5 has the bandwidth to
  3754. run 1000baseT.  But early experience with 1000baseT showed that 1000baseT
  3755. was pickier about certain cabling issues that weren&apos;t specified in the
  3756. Cat5 standard, such as crosstalk and delay skew, so the Cat5 standard was
  3757. enhanced for 1000baseT to enforce limits on these.  This enhanced standard
  3758. is called Cat5e.  But the fact is that most Cat5 installations already
  3759. perform to the Cat5e spec.
  3760.  
  3761. &lt;p&gt;If someone tells you to rip out a Cat5 installation because it can&apos;t
  3762. support 1000baseT, you&apos;re being prompted to do something that is expensive
  3763. and probably unnecessary.   All you generally need is test the existing
  3764. cables to the Cat5e standard (using a Cat5e cable tester) and replace
  3765. the ones that fail.  Often, most if not all the cables will be fine.
  3766. Or just use the cables for 1000baseT and replace any that exhibit problems.
  3767.  
  3768. &lt;p&gt;Cat6 and Cat6a are a different matter.  Cat6 supports a spectral
  3769. bandwidth of 250MHz, up from Cat5/Cat5e&apos;s 100Mhz, while Cat6a supports
  3770. 500Mhz.  Cat6 cabling will run ten gigabit ethernet (10GbaseT) to 37-55m,
  3771. while Cat6a will run 10GbaseT to 100m.  So it&apos;s worth choosing Cat6 or
  3772. Cat6a over Cat5e for new cabling, if the cost increment isn&apos;t too high,
  3773. so that the cabling can support 10GbaseT, even if it&apos;s not needed today.
  3774. </description>
  3775.  </item>
  3776.  <item>
  3777.    <title>Einstein&apos;s special relativity isn&apos;t as complicated as many people seem to think.</title>
  3778.    <pubDate>Mon, 30 May 2011 21:26:00 -0400</pubDate>
  3779.    <link>http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2011/05/30#relativity</link>
  3780.    <category>/misc</category>
  3781.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/misc/relativity</guid>
  3782.    <description>
  3783. &lt;p&gt;I run into people who think that special relativity is some sort
  3784. of mysterious thing that only Einstein and physicists can understand.
  3785. But it&apos;s not.  It&apos;s a bit weird, but it&apos;s no weirder than the earth being
  3786. a globe.
  3787.  
  3788. &lt;p&gt;Originally people thought that light moved like any other moving
  3789. object.  Einstein thought about this and wondered: what would happen if
  3790. you followed some light and sped up until you travelled at the same speed
  3791. as it.  Then light would look to you like it was stopped.  But stopped
  3792. light (light &quot;standing still&quot;) didn&apos;t (and still doesn&apos;t) make sense.
  3793. So Einstein thought: what if light travels at the same speed no matter
  3794. how fast you&apos;re going?  What would this mean?
  3795.  
  3796. &lt;p&gt;Well, what does it mean to travel &quot;at the same speed&quot;?  It means light
  3797. covers the same amount of distance in a given amount of time.  Or, put
  3798. another way, light takes the same amount of time to cover a given distance.
  3799. So if the distance is short, light takes less time to go the distance.
  3800. If the distance is longer, light takes proportionally more time to cover it.
  3801.  
  3802. &lt;p&gt;So Einstein thought: OK, if light travels at the same speed for everyone
  3803. no matter how fast they&apos;re going, what would that mean for someone going
  3804. very fast?  Imagine they&apos;re going nearly the speed of light, and are being
  3805. chased by a beam of light.  Clearly the light isn&apos;t going to get closer to
  3806. that person as quickly as it would get closer to someone who was standing
  3807. still.  Ordinarily, you would think that light was moving &quot;slower&quot; for the
  3808. person who is moving away from it.  But if light moves at the same speed
  3809. for everyone, than something else must be going &quot;slower&quot; for that person.
  3810. The only possibility is time.
  3811.  
  3812. &lt;p&gt;Put it this way: light covers a certain distance in a second.  To someone
  3813. watching, the pursuing light isn&apos;t making up the distance quite so fast
  3814. between it and the moving person, because the person is moving away so fast.
  3815. But for the moving person, light is moving as fast as it always does,
  3816. it is the second that takes longer.
  3817.  
  3818. &lt;p&gt;This sounds a little bit crazy since we aren&apos;t used to thinking of
  3819. time moving faster for some people and slower for others.  But it does.
  3820. The reason we don&apos;t notice is that the speed of light is very fast and we
  3821. can&apos;t easily go at speeds close to it.
  3822.  
  3823. &lt;p&gt;It&apos;s the same sort of thing as the world being round (i.e. a globe).
  3824. It looks flat to us, but only because it is so big that we can&apos;t see enough
  3825. of it at once to see it curve.  Go high enough and we can see the curve
  3826. of the earth&apos;s surface easily enough.  
  3827.  
  3828. &lt;p&gt;Similarly with special relativity.  Time moves slower for those who
  3829. move fast.   It&apos;s not obvious to us because we usually don&apos;t move very fast,
  3830. so at the speeds we move, the time differences are too small to
  3831. notice.  But in 1971, &lt;A HREF=&quot;http://www.time.com/time/magazine/article/0,9171,910284,00.html&quot;&gt;Joseph Hafele and Richard Keating took some
  3832. very accurate (cesium atomic) clocks abord commercial airliners and flew
  3833. around the world&lt;/A&gt;. They compared their clocks to the very accurate
  3834. clocks in the US naval observatory: the clocks were indeed different,
  3835. and &lt;A HREF=&quot;http://www.sciencemag.org/content/177/4044/168.short&quot;&gt; showed
  3836. the results that Einstein had predicted. &lt;/A&gt;
  3837.  
  3838. &lt;p&gt;What this this mean?  Well, if you can wrap your head around the concept
  3839. of the world being a globe, you can wrap your head around the concept of
  3840. time moving more slowly for those who move fast.  And that&apos;s it, right?
  3841.  
  3842. &lt;p&gt; Well, not really.  There&apos;s also general relativity (and it affects
  3843. Hafele and Keating&apos;s results too).  But that&apos;s a bit more complicated,
  3844. and I&apos;m not going to get into it now.
  3845. </description>
  3846.  </item>
  3847.  <item>
  3848.    <title>Exchanging files in docx format may lead to problems</title>
  3849.    <pubDate>Wed, 23 Feb 2011 11:10:00 -0500</pubDate>
  3850.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2011/02/23#docx</link>
  3851.    <category>/it</category>
  3852.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/docx</guid>
  3853.    <description>When Microsoft came out with Office 2007, the default save format for
  3854. files was switched to a new format based on XML.  For Microsoft Word,
  3855. for example, instead of files being saved in &lt;em&gt;.doc&lt;/em&gt; format by default,
  3856. they were now saved in &lt;em&gt;.docx&lt;/em&gt; format.  If you use Microsoft Word 2007 or
  3857. 2010, you&apos;ll notice that when you save a Word document, it saves
  3858. it as &lt;b&gt;document.docx&lt;/b&gt; instead of &lt;b&gt;document.doc&lt;/b&gt;.
  3859.  
  3860. &lt;p&gt;Unfortunately, now there seems to be an incompatibility between how
  3861. Word 2007 and Word 2010 interpret &lt;em&gt;.docx&lt;/em&gt; files.  Apparently,
  3862. possibly depending on how one&apos;s printer is configured, when users of
  3863. Word 2007 and Word 2010 share files in &lt;em&gt;.docx format&lt;/em&gt;, some spaces
  3864. (seemingly random) between words in the file are dropped.
  3865.  
  3866. &lt;p&gt;This has been reported on various places on the net including the
  3867. &lt;A HREF=&quot;http://www.bnet.com/blog/businesstips/word-2010-spacing-problem-solution-actually-just-a-workaround/8049&quot;&gt;CBS Interactive Business Network&lt;/A&gt;,
  3868. &lt;A HREF=&quot;http://news.cnet.com/8301-1001_3-20034213-92.html?tag=nl.e724&quot;&gt;cNET.com&lt;/A&gt;, and &lt;A HREF=&quot;http://social.answers.microsoft.com/Forums/en-US/wordshare/thread/2764c5ac-4f7c-4a6d-9419-9e37bddf82d8&quot;&gt;Microsoft&apos;s own user forums&lt;/A&gt;.
  3869.  
  3870. &lt;p&gt;For now, I suggest using the older &lt;em&gt;.doc&lt;/em&gt; format for
  3871. users of different versions of Microsoft Word to exchange
  3872. documents.  For publishing documents, instead of using a native Word
  3873. format, I suggest using a widely-used open document standard like &lt;A
  3874. HREF=&quot;https://secure.wikimedia.org/wikipedia/en/wiki/Portable_Document_Format&quot;&gt;PDF&lt;/A&gt;.
  3875. &lt;A HREF=&quot;http://www.cutepdf.com/&quot;&gt;CutePDF&lt;/A&gt; is a useful free Windows
  3876. printer driver that lets you create PDF files from any Windows application
  3877. by simply printing to a &lt;em&gt;CutePDF&lt;/em&gt; printer.
  3878. </description>
  3879.  </item>
  3880.  </channel>
  3881. </rss>
  3882.  

If you would like to create a banner that links to this page (i.e. this validation result), do the following:

  1. Download the "valid RSS" banner.

  2. Upload the image to your own server. (This step is important. Please do not link directly to the image on this server.)

  3. Add this HTML to your page (change the image src attribute if necessary):

If you would like to create a text link instead, here is the URL you can use:

http://www.feedvalidator.org/check.cgi?url=http%3A//www.cs.toronto.edu/%257Ejdd/blog.cgi/index.rss

Home · About · News · Docs · Terms

 
Copyright © 2002-9 Sam Ruby, Mark Pilgrim, Joseph Walton, and Phil Ringnalda