It's oddly appropriate to read about this now, as I've just spent two weeks switching from Windows to Ubuntu. You definitely get a feel that the industry favors Windows as a platform, especially if you want to spend all your time in a GUI, but they're both very workable. With Wine and virtualization (Virtualbox, VMWare, and others are available free) I've managed to get everything I've attempted at least up and running. As of now I'm only using Wine for are the heap of small one off tools I've collected for various calculations (MiscEl, guitar tone stacks and the like), and I'm using a Virtualbox XP guest for the IAR Studio that came with the Launchpad just to tide me over. I got the mspGCC toolchain based on GCC4 up and running, and I'll probably switch to that after I get more comfortable with it. Qucs is awesome, though I haven't had much time to really play with it yet, and Kicad even looks better in it's native environment. I'm sure enterprise level tools would be a different story, but I'm blessed with poverty and hobby level interest.
I'd never heard of SILVACO before, and looking at their site it's definitely enterprise level stuff. It struck me as odd that something like that would require virtualization for use under windows so I went poking around to see what the story was. Their windows compatibility looks troubled, and it's easy to see why it'd be better to use a different platform.
I wonder why the change, maybe virtualization is easy enough anymore that they don't want to try to maintain multiple platforms? It'd be interesting if that were a trend for enterprise level software. I can see it being a popular idea as it cuts their costs and leaves them free to work on their product. I've got no idea really, just guessing.
I thing gEDA more or less exists as a tool that interested parties can pick up, use if it suits them, and hopefully sharpen for others. The *nix dependence is a huge turnoff for many, agreed, and for a simple hobby tool like Kicad (which I imagine is largely used in it's windows form) it would have made it relatively obscure. As it is, gEDA strikes me as an advanced hobby tool that gains a lot of it's strength from having a plain text file format that's post-process friendly. This is candy to someone with a programming background but relatively irrelevant to those used to dealing with a program at face value. There are other benefits and drawbacks to be sure, and it's evolving. It's on my list of things to tinker with after I'm more settled in to my new home.
To each their own, naturally. I do hope gEDA evolves into a cross platform EDA, hell, I hope it merges with Kicad and they take the best of both to build on.
Sorry for all the rambling (was really bad this time). Just my two cents.