I doubt its going to be pushing "develop for linux" that will be a problem, most games aren't using many os calls and those are generally easily ported anyway.

The challenge will be getting developers to program for opengl/sdl/openal instead of directx, something they could already do on windows but have chosen to go the directx route. The mobile push has opened a lot of developers up to opengl ES and opengl in general sucks less now so this may not be as difficult a pitch as it would have been 5 years ago.

What would be really cool, is if valve decided to throw their weight behind building a direct 3d state machine for linux, I know there is work underway there but I would imagine valve could seriously speed that up along with being in a position to apply pressure to the video card manufacturers for support in a way that the OSS community simply can't.

Heh, interesting - Google's ANGLE project supports OpenGL ES on top of D3D 9, and if another project implemented D3D on top of OpenGL... would it ever matter which you used?

The linux attempt at this in Gallium3D[0], which is the previous work I was alluding to in my last post. I believe a D3D 10/11 state tracker exists for it but the problem is that only the open source ATI and Nvidia drivers support it. Unfortunately the 3D performance of those drivers is somewhere between broken and terrible so it doesn't mean much today.

> if another project implemented D3D on top of OpenGL

This is exactly what wine does.

[0] http://en.wikipedia.org/wiki/Gallium3D

This is where Valve blows the doors off the world, and announces they've ported the DirectX API to Linux

I do imagine thats a huge part of it. Combine that with the fact that if you target Valve's hardware, you're also targeting a pretty decent sized cross section of Linux PC's as well, and you've got yourself a recipe for better than what the competition has (from a developers perspective)

Developing for SteamBox is a lot less scary than developing for "Linux".

Linux has even more hardware diversity than Windows, and the added complexity of multiple distributions and competing drivers (e.g. open source vs closed). And fans that will bite your head of if you don't support their favorite choices.

If Valve does it right, the SteamBox will define the baseline for gaming on Linux. Distributions targeted to home users will be sure to include the ABI's used by the SteamBox, and Linux gamers will buy hardware that is "similar, but better" than the SteamBox hardware to maximize the chance their Steam library will work on their PC.