The discussion was pretty much “oh, that’s an interesting metric” but everyone pretty much agreed with what you say, there is no way to get a very accurate picture.

In addition to what you cite, there are also huge fleets of Linux servers out there which don’t do web browsing at all (so wouldn’t be tracked here) and you have corporate and educational deployments where lots of individual machines come out just a handful of IP addresses… the list goes on :)

I’ve seen a lot of time digesting those stats for my own reasons as they are to date the only public data source I am aware of which a credit (and publicly reviewable) collection methodology. And I’d like to get an idea of how your community is approaching analysis of these numbers.

Some important caveats to note which make using those stats for Linux vendor specific analysis very difficult.

Mint 11 and 12 report as Ubuntu in the useragent string. You can not untangle Ubuntu usage from Mint usage in those stats.

Mint Linux Debian Edition reports as vendor unknown (as does Debian and Fedora) The unknown linux category is a major problem for vendor specific analysis. As of April’s stats, the unknown linux is larger than _all_ vendor-specific linux taken together (not including Android.)

Essentially… its impossible to get a statistically valid picture of any linux vendor usage( other than Android) from those stats. The unknown vendor usage is just too large to say anything of merit about one linux vendor relative to another, even Ubuntu. At best you can trend over all linux adoption (with and without Android) across a span of a couple of years.

-jef