Monthly Archives: October 2008

Fun page on statistics

I was looking for a reference to the story about bullet holes in bomb planes, and came across this fun page on statistical lore. My favorite:

Question: How many people have more legs than the average?
Answer: Almost everyone. This is because the number of three-legged people are greatly outnumbered by one-legged people, so the mean (i.e. the posh mathematical way of saying that which most people think of as the ‘average’ [total sum divided by number of values]) number of legs is a little bit lower than 2.

I also liked the fact that, statistically speaking, there are 2 popes per square kilometer in the Vatican….

Plus ca change….

I clipped this from ACM Technews, an email service of the ACM:

Looking for Job Security? Try Cobol
IDG News Service (10/23/08) Sullivan, Tom
A Cobol programmer may be one of the most secure and steady jobs in IT. Analysts report that Cobol salaries are rising due to a healthy demand for Cobol skills, and there are few offshore Cobol programmers. The troubled economy also bodes well for Cobol programmers, says Interop Systems director of research Jeff Gould, as long as they are working for an organization that intends to keep its legacy Cobol applications. "Many mainframe customers with large mission-critical Cobol apps are locked into the mainframe platform," Gould says. "Often there is no equivalent packaged app, and it proves to be just too expensive to port the legacy Cobol to newer platforms like Intel or AMD servers." Deloitte’s William Conner says salaries for Cobol programmers are rising because many Cobol programmers are reaching retirement age and colleges are focusing on Java, XML, and other modern languages instead of Cobol. Dextrys CEO Brain Keane says Cobol programmers are less likely to have their jobs outsourced because the Chinese do not have mainframe experience and recent Chinese computer science graduates have focused on the latest architectures and systems and do not have experience with legacy languages and systems. Meanwhile, warnings that mainframes would disappear have proven to be untrue, particularly because mainframes are very reliable at handling high-volume transaction processing, and companies are increasingly benefiting from integrating legacy mainframe Cobol applications with the rest of their enterprise.

(Full article here.) With the exception of the part about offshoring, this article could have been written 10 years ago, and be just as true then. There are, of course, a number of programmers in India that know COBOL – converting legacy apps for the year 2000 was one of the jobs that got the Indian IT service industry started.

Come to think of it, I never really learned COBOL, myself. But I was a decent REXX programmer….

Tim O’Reilly nails it on cloud computing

In this long and very interesting post, Tim O’Reilly divides cloud computing into three types: Utility, platform and end-user applications, and underscores that network effects rather than cost advantages will be what drives economics in this area. (This in contrast to the Economist’s piece this week, which places relatively little emphasis on this, instead talking about the simplification of corporate data centers – though the Economist piece is focused on corporate IT.)

Network effects happen when having new users on a platform or service are a benefit to the other users. This benefit can come from platform integration – for instance, if we both share the same service we can do things within that service that may not be possible between services, due to differences in implementation or lack of translating standards.

Another benefit comes when the shared service can leverage individual users’ activities. Google’s Gmail, for instance, has a wonderful spam filter, which is very reliable because it tracks millions of users’ selections on what is spam and what isn’t.

Tim focuses on the network effects of developers, which is an important reason why Microsoft, not Apple, won the microcomputer war. When Steve Ballmer jumped around shouting "developers, developers, developers", he was demonstrating a sound understanding of what made his business take off – and was willing to make a fool of himself to prove it.

Tim also invokes Clay Christensen’s "law of conservation of attractive profits", arguing that as software becomes commoditized, opportunities for profits will spring up in adjacent markets. In other words, someone (Jeff Bezos? Larry and Sergei?) need to start jumping up and down, shouting "mashupers, mashupers, mashupers" or perhaps "interactors, interactors, interactors" and, more importantly, provide a business model for those that build value-added services on top of the widely shared platforms and easily available applications they provide.

One way to do that could be to make available some of the data generated by user activities, which today most of these companies keep closely to themselves.  That will require balancing on a sharp edge between providing useful data, taking care of user privacy, and not giving away your profitability too much. As my colleague Øystein Fjeldstad and I wrote in an article a few years ago – the companies playing in this field will have to make some hard decisions between growing the pie and securing the biggest piece for themselves.

If we cannot harness network effects, cloud computing becomes a cost play, and after awhile about as interesting, in terms of technical evolution, as utilities are now. USA is behind Europe and Asia in mobile phone systems partially because US cellphone companies were late in developing advanced interconnect and roaming agreements, instead trying to herd customers into their own network. Let’s hope the cloud computing companies have learned something from this….

The Economist on cloud computing

The Economist has a nice feature on corporate IT, mostly about how it is evolving towards cloud computing. Like everything the Econonomist does, it is nicely worded, measuredly opinionated, and contains nothing new to those in the know. But the article is a very good introduction to the current state of at least the technical and market side of corporate IT provisioning, so I mark it for future courses on just that topic. It includes an audio interview with Ludwig Siegeler as well.

CACM becomes much more readable

CACM (Communications of the ACM) is one of my favorite journals – and it is currently in the throes of an editorial upheaval that I think is very positive. In addition to scholarly articles, it is moving in the direction of essays and more generally accessible articles, without loosening the quality criteria. Ever since BYTE disappeared (a victim of the need for targeted advertising) I have missed a general, quite technical yet accessible journal – CACM is now getting closer to what I am looking for.

Here are two articles I found very interesting:

  • "Will the Future of Software be Open Source?, a well reasoned reflection by Martin Campbell-Kelly, giving a very terse, yet comprehensive and useful description of the evolution of software markets. Answer: OS is a tempting conclusion if you extrapolate, but extrapolation has not been a very successful prediction technique so far…
  • "Searching the Deep Web", by Alex Wright, which explores two different approaches to searching beyond static web pages – the trawling approach, which relies on local storage, and the angling approach, which produces targeted results in real time.


Brad Templeton has a long and good analysis (containing spoilers) of Neal Stephenson’s Anathem, which I read a couple weeks ago and have yet to make up my mind about. On the one hand, it sets up a great world with Concets of Avout who devote themselves to science rather than Praxis, it invents a number of words and does quite a bit of philosophic reasoning on topics Stephenson has explored before, such as multiple universe models. I like the first 200 pages or so really well. Then it becomes a picaresque, and not a good one at that – similar to the tour around the world in the middle book in the Baroque trilogy. Lastly, it becomes a tad puerile, with people flying around in space suits and boarding spaceships.

I love the language that Stephenson creates, and the notion of scientific communities locked in for either 1, 10, 100 or 1000 years (depending on how dangerous their exploits are, it seems) is very interesting. But the plot line could do with some sharpening, and the character descriptions are shallow at best. As is usual with Stephenson, mind you.

So, make up your own mind. I still think Cryptonomicon is Stephenson’s best, but maybe that is just me.

(Minor quibble: I think I found an error, and am enough of a nerd to report it. On page 512-13, we find the sentence "Late yesterday, Yul had shattered the calm by starting the engine of Cord’s fetch, and …." But Cord’s fetch was left on the other side of the pole, wasn’t it (on page 416)? Oh well…..maybe I should update the Anathem Wiki. On the other hand, I have a life.)

Incidentally, Anathem may be the only book published so far that has its own video trailer without first being made into a film. Here it is:

The World of Anathem,t=1,mt=video

Airport insecurity

I am thinking a lot about security now, since a discussion last week on security in the 2.0 Enterprise – where the conclusion was that we need to get away from perimeter security and over towards something asset-based, i.e. securing what really matters and not faking security by having showy and inconvenient moats and drawbridges.

This funny but deeply serious article in The Atlantic takes on the example of airport security with all its symbols and holes. As Bruce Schneier (a real security expert) repeatedly has pointed out, hijackers can no longer get into the cockpit. Furthermore, passengers would attack hijackers on sight, rather than cooperate with them. Hence, the bluff that got the 9/11 hijackers in control of four airplanes will no longer work.

But we persist in implementing security that does little but increase the cost of flying, inconveniencing everyone, and, ironically, making flying (or, at least, turning up at the airport) less secure. As the article points out, the most dangerous place in the airport is where many people are waiting closely together in an unsecured area. In other words, in the security control line, perfect in case somebody wants to repeat the Lod airport massacre.