Is Netcraft an accomplice or merely a victim?
“It’s part of a continuing behavior pattern by Microsoft that I think it’s fair to call “dirty fighting.” GoDaddy was using Apache (I assume on Linux) because it was a great technical solution. They didn’t switch to IIS on Windows Server 2003 for any technical reason. The switch was accompanied by a press release by GoDaddy, containing Microsoft promotional language. Now, I’ve changed many servers from one thing to another, but I’ve never made a press release about it. GoDaddy wouldn’t be doing that unless Microsoft had offered them something valuable in return. There has been talk in the domain business that Microsoft has been offering the large domain registries a wad of cash to switch their parked sites. There is no other reason to do this than to influence the Netcraft figures.”
[T]he GoDaddy, Google and other Netcraft debacles in the Spring of 2005 and later revealed that Netcraft does everything it can to twist the data in such a way that it increases Microsoft’s webserver percentages and decreases FOSS’s. With that twisting they have been able to make their “reporting” APPEAR that Apache’s server share has dropped from over 70% to 46%, while Microsoft’s has risen from under 20% to nearly 30%.
Deep down in their web site Netcraft concedes that in current “heavily used” websites Apache runs 66% while Microsoft runs only 18%. (See below!) In other words, for example, Netcraft counts the HUGE number of idle Microsoft servers that GoDaddy PARKS, while NOT counting ALL the web servers in Google’s ACTIVE Linux server farm. Their logic for this hypocrisy is mind numbing. They use other similar “metrics” to extend the same distortions.
By comparison, another site which tracks web server market share is SecuritySpace:
which shows Apache’s current share at 71% and Microsoft’s is at 17%. From a dropdown combobox on the same page you can query various domains and countries to see how FOSS and MS compare in percentages. For example, France is 89% FOSS and 8% MS.
As noted before in the context of Apache, Microsoft may be exploiting the nature of Netcraft statistics to promote itself, so this might not be Netcraft’s fault. Nonetheless, it puts in serious doubt any output from Netcraft because its numbers were essentially ‘hacked’ in the sense that they are now being gamed. With regards to uptime, another person points out:
Apparently in both cases, Microsoft offered to the customer “twice the quantity of servers (that normally would be needed in a RedHat/Apache implementation), and twice as many systems admins with full redundant service, “all at Microsoft’s expense”. The aim was to publicly show “reliability, scalability” of Microsoft technology as compared to the more established and respected competition.
We’ve already seen just how unreliable Microsoft's poster child is. It’s said to be another case of heavy redundancy for the illusion of stability.
To conclude, says another commenter:
I had the same experience as you, GreyGeek. When I found out that Netcraft lost its objectivity (even its appearance of objectivity), and started rigging the numbers in late 2005 to make Microsoft’s net presence look better than it actually was, I lost all respect for Netcraft’s web survey.
Respect for Netcraft is so low now that it might as well just start reporting IIS as having 100% market share across the board. It wouldn’t change anyone’s opinion of Netcraft, so no harm could come from it. The only difference is that Netcraft could stop pretending it isn’t an industry joke.
For what it’s worth, according to Netcraft, Boycott Novell is the 2156th Web site in the world, as ranked by traffic. Although we served over 30,000 pages/day earlier in the week (as judged by AWStats metrics), it’s very doubtful that Netcraft’s traffic rank is not skewed in favour of system administrators. For similar but inversed reasons, Alexa ranks are GNU/Linux- and Firefox-hostile (more Windows/IE-oriented). I wrote about this in my personal Web site, which was at some stage ranked around 17,000th, according to Alexa.