There's also a big difference in the user experience: users are usually far less senstive to the OS or server software than they are to a new brand (or even just a new version) of their desktop software. And, much as I like OpenOffice myself, it's difficult to argue that Microsoft hasn't learned some lessons and focussed on better versions of its office software.
When it comes to databases, I'm pretty sure the 80/20 rule would apply: 80% of any companies databases could run on pretty much any RDBMS; the other 20% might take some, or even a lot, of work.
Biting the hand that feeds IT
