Debian also packages the driver, but you still can install it "Nvidia way" which allows installing the newer ones which are ahead of what's packaged. sgfxi is a great tool which automates it all: http://smxi.org
Steam numbers are quite useless as is to estimate Linux sales potential. What you should evaluate are percentages of sales per OS for cross platform games (for example ones which release all versions for all OSes at once). Steam doesn't offer such statistics. Also don't forget that a lot of Linux users avoid DRM and Steam as well as a result. So to have a better estimation you should get such stats for games which are DRM-free. May be GOG can publish some?

They didn't even start selling them, so there is nothing to evaluate yet.

I think that when it comes to game creation, most developers don't deal with the DirectX/OpenGL level. Engines make creating graphics pretty seamless. In this respect I think that Linux gaming would continue to grow, as long as people there are paying for games, because there's less need to deal with the platform differences if you're using Unity or Unreal Engine, and most devs will go that way.

If developers have no experience with Linux, the fact that engine supports it won't help them. You can expect to get a junk version in which developers won't be fixing any bugs because they can't program for Linux. May be for lower end games it's less of a problem, but for high end ones it's for sure more important.

I guess you're right, but I hope that won't stop Linux gamers from buying these games (and providing suitable feedback). The more people but games for Linux, the more developers will see it as a target market, and the more experience they'll get with it.

That may be, but it can also create a reputation of Linux versions being always broken. More than once I saw Unity based games made by developers with little knowledge of Linux who just press some "export" button and don't care to address any bugs reported to them because they probably even can't.

More and more people are building their own system because it's actually quite easy to do now. A friend of mine who is passionate console gamer once said that PC's are big bulky and stupid. After building a custom rig her opinion has changed.

May be more, but I strongly doubt it's a majority of users.

I'll have to see some stats to believe that the number of people building custom PC's is growing.

The industry has a habit of saying one thing and doing another. I am reminded of Jim Sterling's coffee episode, when he pointed that among people served everyone said they prefer their coffee black when in fact most people prefer it with milk. Point is people said they prefer black because it's trendy thing to say.

The games industry can stop using Direct X anytime they want. It doesn't even seem like it would even take a lot of effort to port something from direct X to OpenGL since the Linux community seems to do a half way decent job for free. That's not true at all. AS I said before Id's John Carmac make the Quack 3 engine OpenGL based and Q3 was one of the most used game engines in the early 00s. So this has been done by people a lot more influential them Valve. Steam is influential yes, but I don't think Valve are as big as they think they are.
You could post a poll on a general gaming site. I can tell you now though that more people build their systems today then they did years ago. But you can also get pre built PC's like the NUC and Brix. So there's still no need for a Steam box.

I'm not sure where I could post it and what to ask to make it feel like I'm getting real results.

It's possible though that a higher percentage of desktop users are building PC's, simply because many of those who don't have moved to laptops.

That's a very long time ago when it comes to APIs. What happened in 2000s is now not really relevant. What's happening right now is very different.

I'd expect a small percentage to build their own systems. I do so myself, but the amount of effort and research it requires is way above the habits of your average user. Enthusiasts do that, yes. But they are way not the majority.

Can someone please explain to me in a very basic way what this new rewrite of OpenGL is about? I have some experience with OpenGL, I was dabbling in it a bit last year, so you don't have to completely avoid technical terms. I know that OpenGL 3 was all about removing the fixed-function pipeline in favour of the programmable shader-based pipeline. So what is the next big leap?

I agree (which is why I doubt the assertion that more people are doing it now), but if most users are moving to laptops, those left using desktops are more likely to be those who build them themselves. So if you're looking at just the desktop market, you might see a higher percentage of custom built PC's. Perhaps still a minority, but a larger one than before.

I think that's a rather poor example to prove your point, as that is an example of trying to use/run a Windows-only game on an unsupported (Linux) system. I understood the current installers (with password-protected RAR files) cause extra problems running the games in e.g. Wine.

I don't consider it different than e.g. a Windows user wanting to run his old (DOS) PC game in DOSBox or ScummVM, by using only the data files from the game. Or me last week wanting to "extract" data files from the original Quake and Quake 2, in order to use them with mods like Quake Darkplaces or KMQuake2 (running them on Windows 7). Yeah, I guess I was being "unorthodox", since I didn't simply try to install and play the vanilla Quake or Quake 2 on Windows 7.

Heck, with many GOG DOS games I've performed extra measures to change the music source from Soundblaster to Roland MT-32 or General MIDI, in order to get better music. Sometimes that has been a bit tricky because GOG has removed the sound setup executables, and even when they are there, it is not straightforward to get to a DOSBox session in order to change the sound card.

One complaint I quite often also hear is that there are too many user interfaces for different Linux distros, and that makes it too complicated to many users. I think that is hogwash, as overall the differences are quite trivial (the "start menu" items arranged in a different order... so what), and generally these changes are much smaller than e.g. the UI changes between Windows 7 and 8, or even 8 and 8.1. If normal users can live with how the user interface can change even drastically between Windows versions, I don't think they should have any problems learning their way around in a slightly different Linux UI.

At home I mostly use XFCE in Linux because I like the idea of it being a lightweight UI (esp. as I quite often run Linux on older PCs with obsolete Windows versions, even with 32bit CPUs), but I have no problem occasionally using Linux with some totally different desktop environment. Ok, the mobile-friendly Ubuntu UI was a bit awkward to me, the same way like Windows 8 Metro was.