Go Daddy moved 4.5 million parked domains from Linux to Windows, which gave Microsoft's Windows Server 2003 a gain of 5 percent market share in the survey's results. Some criticized Netcraft for measuring results according to IP addresses rather than physical servers, claiming it offered an inflated representation of server market share.

Netcraft responded that its methodology was well documented, and that its hostnames-based report was a specific and valuable metric.

Leading the group of critics was active open source evangelist Bruce Perens, who accused Microsoft of bribing Go Daddy to move the parked domains to Windows-based servers. Seeking to rectify the situation, Perens created an open source initiative entitled Open Source Parking (opensourceparking.org).

The project set out to increase the market share for open source software, as well as generate advertising revenue for the open source community by encouraging domain registrars and owners to park their unused domains on Linux-powered Apache boxes.

But generating revenue for open source causes proved to be harder than Perens anticipated since Google for Domains refused to help, even after Google's open source evangelist Chris DiBona intervened.

"I tried some other domain parking advertising vendors," says Perens. "Some were slimy and had nasty multiple pop-ups, browser bombs, even spyware. Some wouldn't take the account because I didn't own the parked domains. Some got more fraudulent traffic than real clicks. So, right now the sites only have some banner ads for OpenOffice and Firefox. I'd be glad to hear from a parked domain advertising broker that could help."

Several weeks ago, Netcraft discovered that Perens had altered server headers in order to improve the survey results of the Apache Web server. Sites hosted on the Open Source Parking service specify that they are running on Apache, though in actuality they are hosted on lighttpd.

The headers read Apache, but contain the Date & Server headers last, which is like a lighttpd response, but unlike a normal Apache response. Additionally, the etag was also found to not be in Apache format, and instead, matched the lighttpd format.

Perens admitted to the claims after Netcraft reported the findings on its page, though he denies any malicious intent on his part. Rather, Perens says he altered the header to indicate the server was Apache, and include the message "; Hello Netcraft!" to "comment on what Netcraft was allowing its statistics to be used for."  Perens says some 3,000 parked sites sent the message to the Netcraft report.

"I think the biggest thing I did was stop people from using the Netcraft survey to convince naive managers that IIS was somehow gaining market share, when it was really only artificially gaining a group of parked sites," says Perens. "It's easy to lie with statistics. The most you can do when somebody does so is point it out as loudly as you can."

Although Apache and lighttpd are both open source Web servers, Apache leads in the Netcraft survey with a 59 percent market share, while the relatively new lighttpd holds just a 1.2 percent market share.

Perens says he wanted to use Lighttpd instead of Apache, but at the time Netcraft was scoring lighttpd as "other," which would not have indicated that he was using an open source Web server.

"Apache is desirable if you want mod_perl, or mod_some-interpretive-language," says Perens. "That's an old-fashioned way of programming. Most newly-architected Web sites decouple the dispatcher running the interpretive language code from the Web server. These days, something like lighttpd works better for most sites. Open Source is about evolution."