Philip Greenspun’s Weblog

We have completed our three-day intensive RDBMS/SQL course at MIT. The feedback from students so far is that they thought that it was more effective per hour of their time than struggling through problems at home. This makes a lot of sense to me since anyone learning a new set of tools can get stuck for hours on something trivial, e.g., not knowing the right keyboard commands or not knowing how to find the error log or not knowing how to invoke or quit the debugger. In a shared lab with other students and TAs, there is a much greater chance of a student getting unstuck before wasting a lot of time.

Students liked using Google Docs and they liked having their own VM.

We were concerned that the Android application development portion of the class would be seen as pointless and far too confusing with the mixture of tools required and the three .java files in Eclipse necessary to do even the most trivial thing (grab XML page from server and display values on the Android screen; if programming is like this, no wonder that all of the smart people in the U.S. go to med school and Wall Street). However, the way that we set it up so that we told them exactly which files to touch and how they ended up liking it. Now they can see all of the moving parts in a collaborative server/db-backed Android application. (We could not do the same thing for iPhone because the development environment runs only on Macintosh, which just a handful of students had, and because none of us knows how to use a Mac!) We spent about 1.5 hours on Android development and the overall structure of DBMS, HTTPD, firewall, XML transport, XML parsing, display.

People came to the course with varying levels of experience, which resulted in some people falling behind. By the time that 70 percent of the class had solved the problem, we had to move on to discuss the solution. It is incredibly efficient to be in one room together, but at the same time the only way to train everyone to proficiency would be to have all the learning be self-paced. I don’t have a good solution to this. We probably could have done better with an “extra credit” harder problem following every standard problem. That way the quicker students would have something to do other than catch up on email while the struggling folks were still working on the basics.

The students didn’t do much with PHP, but it proved inoffensive and didn’t waste anyone’s time. Now that MySQL is 16 years old, I was surprised at the things that they’ve yet to accomplish. For example, the C in ACID is “Consistency”, i.e., being able to enforce rules in the SQL data model and using the RDBMS as a last line of defense against programmer mistakes. Yet MySQL silently fails to enforce any CHECK constraint. Similarly, when students would GROUP BY column_a and then SELECT column_b (where column_b had multiple values within each group), instead of raising an error as Oracle, Postgres, or SQL Server would, MySQL happily picks a column_b value at random and includes it. I don’t understand why people use it. Is MySQL/InnoDB faster than PostgreSQL? If so, and if people don’t need such things as CHECK constraints, why not simply use a NoSQL system?

One sober reflection upon MySQL’s capabilities was captured during an instant message exchange among course instructors while developing the problem sets:

• I’m too old to look shit up in the MySQL manual
• fuck these people
• making something that sucks ass compared to Oracle
• and then making us learn all of the differences

Full post, including comments

I’ve drafted an article on Using Google Docs for Classroom Instruction and would appreciate feedback.

Thanks.

Full post, including comments

We survived the first day of our intensive RDBMS programming course at MIT. About 25 students showed up at 10:00 am and all returned after the lunch break. Though the syllabus is new and untested, the students reported that the class was better than their average MIT course and more effective for learning. I think this goes to show that almost anything works better than the traditional lecture format. We had individual students or pairs working together on laptops and three teachers milling about the room offering assistance for solving problems (plus Andrew in California on IM and Skype). We used Google Docs to distribute the assignments and as a shared workspace for doing code review once problems were wrapped up and ready for discussion.

The only thing that did not work well was distributing the 2.7 GB .zip file containing the virtual machine for the course. It took at least 30 minutes to download even though it was parked on web.mit.edu. We had a couple of DVD-ROMs, but should have brought a bunch of USB sticks. The MIT wireless network and/or the Web server is not ready for intensive use. We used virtualbox.org and had sysadmin problems on all underlying platforms (the mix was approximately 80 percent Windows, 10 percent Macintosh, 10 percent Linux; one guy had a $150 Android tablet for viewing documentation). It would have been ideal to get everyone set up well before the class started, though many of the attendees did not pre-register so it is unclear how we would have reached them.

Overall it was a very rewarding experience. A lot of very smart people showed up and we were able to save them a lot of time while teaching them some useful skills. I need to offer a public thank-you to Andrew Grumet,  John Morgan, and Shimon Rura. Given the range of systems that we had to deal with as well as the problems students encountered that were peculiar to their laptops or “just plain peculiar”, there was no way that I could have been effective alone. So thanks, guys!

Full post, including comments

Folks:

We’ve drafted an RDBMS versus NoSQL article for the three-day RDBMS course that we’re teaching this weekend at MIT (you’re welcome to take the class, even if you’re not an MIT student; just email me). The article is intended to support a 20-minute discussion and give students pointers should they be interested in learning more. As neither Andrew nor I are NoSQL experts, comments would be appreciated.

Thanks.

Full post, including comments

As I was apparently the only Jew handy during a blizzard that dumped 14-18″ of snow in the Boston area, I was asked to comment on whether Sarah Palin’s video (youtube) regarding the Tucson shootings was offensive. Here’s my analysis

• 30 seconds: “our exceptional country, so vibrant with ideas” … “a light to the rest of the world”; The rest of the world doesn’t seem to be in a rush to copy our political system. Most democracies are parliamentary. Do we have better or more ideas per capita than other countries? Certainly we’ve had more per-capita wealth for implementing those ideas, but where is the evidence that we’re more creative than everyone else on the planet?
• 60 seconds: “inexcusable … evil man took lives”; is there any evidence that the killer himself wants to be excused? Certainly nobody else has come forward to say “I think what Jared Loughner did was reasonable under the circumstances”
• 1:24: “spent the last few days… praying for guidance”; if God helped out with this video, I would have hoped that He could have varied the camera angle a little (yes, I know that I’ve done some equally lame work for my helicopter stuff, but my production budget is lower than Ms. Palin’s)
• 2:40: “Obama would join me in affirming the health of our democratic process” [based on there being some turnover of seats in every election]; contrary to Ms. Palin’s statement, it is extremely difficult for an incumbent Representative to be voted out of office. Congressmen and women choose their voters by rearranging districts; voters do not choose Congressmen in most cases. If the vulnerability of incumbents is a measure of political health, as Ms. Palin suggests, we’re not in the best condition.
• 3:31: “journalists should not manufacture a blood libel”; analyzed separately below
• 6:20: “it is in the hour when our values are challenged that we must remain resolved to protect those values”; this has been a very common sentiment expressed in Supreme Court opinions, e.g., when FDR was advocating throwing out the Constitution in order to deal with perceived emergencies during the Depression and World War II. The government would say “we need to regulate all of this stuff that has nothing to do with interstate commerce because of the crisis” or “we need to intern more than 100,000 Japanese-Americans without a trial because they might be a risk”. The sentiment is almost always in the dissenting opinion, with the majority agreeing with the government that the crisis trumps whatever freedoms people had formerly thought the Constitution gave them.
• 6:30 “we had to fight the tendency to trade our freedoms for perceived security [after 9/11]”; Did we win this fight while I was waiting in the security line at Logan for my new X-ray scan?
• 6:55: “we need God’s guidance”; this seems like a bad way to deal with criminal justice or politics since people of different religions will get different guidance from God, at least to judge by what is written down in the world’s various sacred texts. A Jain, for example, would surely not hear divine voices telling him to invade another country, execute a convicted criminal, or do a lot of the other stuff that our government does.
• end of video impression: Palin’s face looks remarkably square (maybe it always looks like this? I haven’t seen any of her other videos and did not watch any TV coverage of the 2008 campaign (since I predicted Obama’s victory back on December 12, 2007))

Was Palin’s use of “blood libel” offensive to me or other Jews? Was it a reasonable analogy? Wikipedia notes that an unexplained murder in Norwich was blamed on the Jews. Palin was discussing an unexplained group of murders (probably nobody other than Jared Loughner will ever know the real reason) and how some journalists were blaming it on people who’ve made statements opposing Democratic Party initiatives.

Personal conclusion: Not offensive and perhaps the most accurate and sensible part of the video.

[Separately, I’m trying not to look at any news about this incident because it would be too sad to think about a 9-year-old girl dying, even if she were the only victim. According to some historians, Stephen Foster wrote Gentle Annie (lyrics; Marilyn Horne singing) after the death of a young girl in a carriage accident. For those of us who lack such talent, what purpose can be served by reflecting on the sad events of last weekend?]

Full post, including comments

Folks: This weekend I’m teaching a class at MIT on RDBMS programming. Sadly I forgot to learn the material myself before offering to teach it. In giving students guidance as to when a standard RDBMS is likely to fall over and require investigation of parallel, clustered, distributed, or NoSQL approaches, I’d like to know roughly how many updates per second a standard RDBMS can process. Here are the criteria:

• mixture of inserts, updates, and selects, with the server doing 10 queries for every update and the typical update or insert involving two rows
• a 4U server with 8 CPU cores and enough RAM to hold the entire database
• eight hard drives so that transaction logs, indices, and tables can all reside on separate spindles (or magic RAID approaches can be used)
• perhaps a little bit of magic in the form of a solid-state drive being used for the transaction logs
• any RDBMS, including MySQL/InnoDB, Oracle, Sql Server, DB2, etc.

I’m looking for a number that is good to within one order of magnitude.

Thanks.

[http://www.tpc.org/tpce/results/tpce_perf_results.asp makes it look as though the answer is “about 2000” though I’m not sure if the TPC-E benchmark simultaneously requires some queries. It is tough to say because the tpc.org Web site is unusably slow…. I finally got the Word doc. It looks like they count some queries as “transactions”, e.g., “The [Customer-Position] transaction is a real-only transaction”. So maybe the answer is closer to 1000? The TPS-E benchmarks use some ridiculously expensive systems, but they also require a huge amount of data to be kept, e.g., 8 TB for a low-end system doing 2000 queries and updates/second.]

(The venerable TPC-C standard never struck me as matching up that well to typical Internet applications.)]

Full post, including comments

One of my big mistakes as an investor was not seeing that a U.S. house price bubble deflation could affect the price of industrial equities (my main investment; I had faith in GE’s ability to sell jet engines in China and India even though I thought houses in Cincinnati were overvalued). Clearing off my desk from 2010, here’s a story from the Wall Street Journal (“Shareholders hit the roof over relocation subsidies”, October 25, 2010) about how whenever top managers at public companies change job, they’re compensated for any loss they might incur from selling their house. The Microsoft shareholders had to pay Stephen Elop $5.5 million in relocation costs, including $3.7 million to make up for the loss on his Los Altos mansion (separately, I would have loved to have gotten the $1.8 million furniture moving contract!).

How much will this take out of shareholders’ pockets? Let’s assume 5000 substantial public companies (e.g., the Wilshire 5000). Let’s assume that each company has 10 top executives, that turnover occurs every five years (2 execs per year), and that each one gets an average of $1 million if he or she sells a house in a down market. Peter Schiff in the WSJ on December 30, 2010 said they’d have to decline another 20 percent to get to the historical trend line (and Philip says the trend line itself is too high because it factors in continued robust growth for the U.S. economy, whereas in fact there has been little growth since 2007); the Dallas Federal Reserve Bank economists say another 23 percent drop is due. So let’s assume the down market in real estate lasts at least another 5 years. The Case-Shiller index peaked in 2006, so that means a total of 10 years of decline. That’s $2 million per public company per year and a total of $100 billion that will go to employees rather than shareholders.

Full post, including comments

The figure above (click for larger version) is from an NSF study of K-12 math and science education and shows eighth graders’ math proficiency as a function of their mothers’ education level. I attended a presentation recently where the figure was used to show that schools should not be held responsible for the math ignorance of their graduates because kids with uneducated mothers were inevitably doomed to incompetence. My first thought upon seeing the figure was “Wow, the schools don’t teach any math at all; kids learn all of their math from Mom.”

Full post, including comments

Readers have been sending me pointers to articles about the Motorola Atrix, a mobile phone with a laptop dock, and saying that it is similar to my 2005 article “Mobile Phone As Home Computer”. I don’t see the similarity because the Motorola dock has no hard drive and isn’t necessarily different than using a bluetooth keyboard and bluetooth display (does that work?) with any other smartphone. My idea of the dock was that it had its own monster CPU and graphics chips, unconstrained by power considerations, plus a hard drive for storing near-infinite video (the one form of data that a typical consumer might accumulate that is too large to be stored inexpensively in the cloud) and a removable hard drive mirror for backup.

Even if Motorola has not implemented my pet idea, this would seem to be a useful tool for travelers. Use the laptop dock when in a hotel or an office; use the phone by itself while walking around. Don’t carry any extra chargers or cables or extra weight from the full hard drive and circuitry of a typical laptop. I don’t like this for replacing a home PC because the 11″ screen on the laptop dock is so small. A $550 Dell home PC includes a 24″ “full HD” monitor with 1920×1080 resolution. That’s enough to view two pages of text side-by-side, a critical ability for students, certainly (one window for Wikipedia, from which to cut and paste into the other window, which holds the term paper).

Thoughts on the Atrix? Anything else from CES strike folks as interesting? There were a lot of tablet announcements, but I didn’t see much in the way of pricing. If a Netbook retails for $300, I would think that a $249 Android tablet would drive consumers truly wild, though this February 2010 analysis of iPad manufacturing cost makes it seem as though a high-performance $249 tablet is at least another year away.

Full post, including comments

I recently rented a 2010 Nissan Altima Hybrid and drove it for about 250 mostly highway miles. When I topped it off for the return to Hertz, I found that it had averaged 28 miles per gallon, i.e., hardly better than the 7-passenger 1993 Dodge Grand Caravan that I owned. The Altima seated four in comfort and five if they didn’t mind getting friendly. The trunk could hold one standard suitcase plus a carry-on bag or two ; much of what would have been the trunk was apparently being used to hold batteries. I learned what 17 years of progress from the automotive industry looks like: going from 25 to 28 mpg while reducing the interior volume of the vehicle by half.

[Typical indicated highway driving speed was 70 mph, with the air conditioner on roughly 60 percent of the time. The trip included about 30 minutes of traffic jams due to an overturned vehicle (how would Southern Californians manage on the icy potholed roads of Massachusetts if they are flipping their vehicles over in broad daylight on perfectly dry and smooth Interstates?) and weekend traffic on a state highway going through what had been a small town and is now part of the endless strip mall sprawl.]

Full post, including comments