The Open Source Census, which I mentioned back in April, just dropped a press release this morning about the data it's been collecting. I chatted the day before with Kim Weins, senior VP of OpenLogic, a key co-sponsor of the census, and how they found a few ... surprises in the results.

Well, maybe they won't be total surprises to people who're intimately involved with the business and culture of open source, but I imagine they'll still raise a few eyebrows. For starters, there is quite a lot of open source software, of all stripes, being deployed on Windows machines.

This includes software deployed as an escape from Microsoft (OpenOffice, etc.), packages used as support or infrastructure tools (zlib, OpenSSL, Perl, Samba), and things that have their own legacy as well -- like Firefox, which appeared on a whopping 84% of the systems surveyed. The vast majority of the top 20 or so packages also appeared on both Windows and Linux systems, so a good deal of what's being used is platform-agnostic.

Another revelation, which probably comes as a surprise only if you haven't been following open source news: Open source adoption in Europe far outpaces that of the United States. I chalked that up to two things: 1) a larger governmental role for open source adoption in Europe, and 2) less existing fidelity toward Microsoft by default there. "Governments and financial service companies" were the biggest leaders as far as use of open source packages, but a chunk who identified themselves as "Other" or "All Others" made up nearly half right there.

What's a little more difficult to determine from just looking at the numbers is the role of the packages used, but I asked Kim about that and learned a few things. Open source is doing great in the infrastructure/component/library space -- for instance, one of the biggest packages used in the financial sector is the all-Java hsqldb -- but it's not quite knocking commercial applications off the desktop yet. For OpenOffice.org, the adoption rate is 73% on personal computers but only 28% on enterprise machines.

The last eye-opener was the breakdown of Linux distributions found in the census. Ubuntu Gutsy and Hardy made up almost half of everything surveyed, with Debian and SUSE at 12% to 13% each. Fedora Core was only 6%, outpaced by -- get this -- Gentoo. Everything else amounted to 12%. (The lack of Red Hat Enterprise here is a hint that the survey's being run mainly on workstations and not servers -- something I hope they're able to.)

The best thing about the Open Source Census is also one of the trickiest things about it: the methodology. By casting the census as a byproduct of an open source discovery tool -- something useful in itself -- that bypasses some of the problems inherent in self-reporting statistics of this kind. The data's at least as useful to the people running the program as it is to OpenLogic and the census organizers.