They’re promoting the survey in response to some gloating by the Apache folks about their 63.98% share in the August 2003 Netcraft Survey.

Port80 looked at the response headers from the home pages of the top 1,000 companies. They found that 53.6% of them ran on some version of Microsoft’s Internet Information Server.

What Are We Counting?

So who is right: Netcraft or Port80? I’ll say neither of them are.

Port80′s correct when they point out that a large number of Apache sites are on shared servers and may be parked, or abandoned domain names. It’s an Apache feature that it’s so easy to use the Virtual Host directive to support hundreds of domains on one server. So the Netcraft survey is flawed in that they don’t distinguish between the importance of amazon.com v. whump.com. Both sites have equal weight in the survey.

However, the 53.6% IIS share of major corporate web sites does not correspond to dominance either. The Port80 survey looks at the parent corporation home site, which may not be the main destination site for that firm’s customers. Or it may be a corporation without a major internet presence that maintains a site for communication with shareholders and the press. So there’s the weighting problem again: cdw.com v. warnco.com. Both run IIS 6.0, but cdw.com is a heavy traffic commerce site and warnco.com redirects to irpage.com

Traffic and Weight

Looking at traffic gets us towards the goal of determining which server platform ‘dominates.’

Nielsen’s site has the top ten destinations by home and business users aggregated to parent company. Again, this is flawed data, as we lose the detail of what application/property people use.

Microsoft, TimeWarner (AOL Server), Yahoo, and Google (Google Web Server) are in the top ten.

Microsoft leads both the home and business use lists, probably by virtue of hotmail.com (IIS 5.0). But how much of that work is done on IIS and how much is still on FreeBSD?

Conclusion

Port80′s right to point out the flaw in interpreting the Netcraft Survey as indicating Apache’s absolute dominance. But they cannot conclude that IIS is the dominant server by looking at raw numbers of the top corporate home pages.

A rigorous analysis must take into account traffic going to every host and what’s running on each host. I’m sure that Nielsen would love to sell you that information if you have means to pay for it.

Aside

Everyone, when you send out numbers or pointers to white papers, please don’t use HTML email. That sets off my “it’s spam” assumption.

During the 1980s Bill James reinvented analysis of baseball statistics. He developed summary statistics to predict the success of players and teams, which, if you play fantasy league baseball, are now standard tools for selecting rosters.

When Beane’s staff in Oakland applied James’ methods to the records of high school and college players: the pool of potential major league players, he found a number who were undervalued. His staff knew this because almost every high school and college team posts their statistics on the Web, where the front office could pull them down into spreadsheets and process them. Many players with high potential for success, based on stats such as the on-base percentage, were ignored by other teams until later draft rounds. Beane could pick up those players in early rounds, and not have to pay a five to ten million dollar premium for them.

Beane already knew, from his experience as a mediocre Major League player, that scouts overvalued many player’s skills. Beane was drafted in 1980 by the Mets and paid a large signing bonus. Ten years later, after several cycles through the Major and Minor leagues, he quit the game and went to work for the A’s front office.

Moneyball: The art of winning an unfair game explains Beane’s success. It’s a great story about Baseball, economics, and the joys of statistics.

L. J. Henderson, a Harvard physiologist and the first president of the History of Science Society, attempted to analyze mammalian blood solely as a physical-chemical substance. He found that the only way he could describe a chemical system as complicated as blood was by a diagram called a “nomogram.”

But as large as his bankroll grew, it only took three weeks to throw it all away. Looking for bigger scores, he started playing a modified Martingale in combination with the Argument From Design. In one terrible streak, Blaise lost everything: His money, his home, his family’s secret recipe for pomme frites. It was a teleological disaster.

I will, however take the certain payoff of $9K. Risk is something you can pay to reduce, and I’m happy to pay $500 in LP’s example.

The Pie is ecumenical, but really, technical analysis? Next thing you tell me is you think “intelligent design” has merit…