The Reg has a long Windows vs Linux report. Linked from /. (discussion), and it seems to take into account things like damage potential, ease of exploitation, size of deployment, busting some myths, etc. I haven't had a chance to read the entire thing yet, but while the Reg isn't the biggest MS fan in the world, I trust their reporting a bit more than Microsoft's "facts" somehow. But hey, I'm biased as well.
I guess the problem is that no matter who does the reporting and comparing they'll have some link to something, or someone will dig up that sometime around 1992 someone in the organization mentioned that "this microsoft thing is kinda cool" and therefor is biased, or they have a linux server so they can't possibly report fairly.
It's also the right tool as Dana is constantly saying, but there are definately crossovers in between linux and windows as far as the tools that are available for both. Anyway, recommended reading of course, and the /. discussion I'm sure will be full of intelligent and calm discussion :)