Inventing Market Share

It irks me that Netcraft still uses "number of domain names hosted on windows vs. linux" as a measurement of windows/linux market share. This is largely apparent from their own analysis of their June 2006 survey.

unique domain name != unique web site

Content that constitutes a single site (or service) is often distributed over multiple domains. Case in point: About.com.

Similarly, content that constitutes different sites is often hosted on a single domain. Case in point (until sometime later today): MSN Spaces.

And then of course, you have the huge numbers of "parked" domain names, which aren't actually being used for anything, but still impact the market share statistics.

The domains/OS metric should be put to rest, in favor of better metrics involving usage, once and for all.

Later today, MSN Spaces blog urls will change from http://spaces.msn.com/<NAME> to http://<NAME>.spaces.msn.com. Dare Obasanjo announced this change on his blog, and made the reasons for the change transparent.

However, I'm fairly certain that within the next 4 weeks, it will be spun as an evil Microsoft effort to tip the Netcraft numbers (see this story, for example). This will happen despite the fact that MSN Spaces was recently declared the most-used blogging service in the world.

The bottom line is that there's probably no evil master plan in play here, and either way, the metric is flawed and needs to be replaced to prevent people on either side from tipping the scale.