NVIDIA's Response To Recent Nouveau Work

[Svartalf]

"Rarely" is still too much. Getting sued only once still can cost millions of dollars. So imho it's still a good reason to keep the source closed.

Actually, a troll will sue you if you remotely look like that you're infringing on their "patent". Seriously. Source code or technical docs will not make it any real easier for them to figure out to sue you over something.

[RobbieAB]

Not wishing to voice the really cynical suggestion, but what if nVidia ARE infringing someones patent? If they even suspect that is the case, the code is staying closed.

I don't see nVidia caring that much what nouveau get up to with their cards, and given the grief AMD/ATI have recieved for "Not opening the blob" since they delivered specs, I wouldn't blame nVidia for not openly helping any FOSS driver. Afterall, damned if they do, damned if they don't, why waste time and resources doing only to be damned anyway.

In the PR stakes we Linux lovers really don't know how to play the game...

[Smorg]

I want to say Nvidia must be insane. The reason this seams like a reasonable course of action is because of the way Andy described the architecture of the proprietary driver. With the vast majority of the codebase being shared between platforms, it stands to reason that they can't simply open-source it for similar reasons AMD couldn't simply re-license their fglrx driver under the GPL.

That sucks for Nvidia, but I suppose they don't really have an alternative unless they tried documenting the hardware and releasing it to open-source developers.

Personally, I'll be buying AMD/ATI from now on. I'm already getting decent initial 3D support on my 4870 with the very first DRI release in 2.6.32 to support it. Its only a matter of time before the quality of the completely free radeon driver blows Nvidia's proprietary driver out of the water. After that happens it will be really hard for desktop users to justify buying Nvidia.

I really don't want to support a company that intentionally distances themselves from those trying to provide free-as-in-freedom support for their hardware.

I don't see nVidia caring that much what nouveau get up to with their cards, and given the grief AMD/ATI have received for "Not opening the blob" since they delivered specs, I wouldn't blame nVidia for not openly helping any FOSS driver. After all, damned if they do, damned if they don't, why waste time and resources doing only to be damned anyway.

It isn't that they didn't want to open their blob, its that they can't. It contains code that AMD doesn't have the rights to release. Even if they did release it, you probably wouldn't want it. Redevelopment in order to cleanly integrate things with xorg seems to be working out nicely. AMD is being fully cooperative when it comes to hardware documentation as far as I can tell.

The important difference is: AMD actually wants free software solutions for their hardware. Nvidia does not.

(1) cause they don't want to share their software solutions with their competitors

Now that is just horrible reasoning. Does anyone honestly think Nvidia is learning anything they don't already know from AMD's code? I highly doubt it. The value added to the product by virtue of FOSS drivers far outweighs any competitive advantage given. If you can build a functional driver through reverse-engineering alone, you aren't going to be giving away any secrets by just opening up the driver. Seriously... anyone who intentionally renders their hardware worthless for that reason can go sodomize themselves with a hot iron poker.

[birdie]

Exactly as I told a few days ago

The important difference is: AMD actually wants free software solutions for their hardware. Nvidia does not.

NVIDIA has nv driver. It's ugly and incomplete but it does support all NVIDIA GPUs and it IS open source. So, let's be frank, NVIDIA just has a lot of third-party luggage in their drivers they'd rather not open.

An, please, don't forget that the world changes and people too. Maybe NVIDIA's just trying to muster its courage to make the first step.

[Xavier]

If nouveau becomes as good as nvidia's binary driver then Nvidia should be very happy about it because Nvidia's only goal is to sell hardware, they keep developing a driver only because without a good driver their hardware would be mostly worthless / much less competitive.

What will most likely happen is that the Nouveau driver will be slightly less good than the binary driver (after all NVIDIA has spent lots of manpower on this), but it will be good enough for nearly everyone to use it for Compiz and everyday stuff. So most people will stop installing binary drivers, which means less bugreports so a lesser quality as time will go. Later NVIDIA will be asked by its big clients to support Nouveau, and that's where they won't be so happy, because they'll have to opensource their advanced techniques in order to do so.

[Apopas]

...

Now that is just horrible reasoning. Does anyone honestly think Nvidia is learning anything they don't already know from AMD's code? I highly doubt it. The value added to the product by virtue of FOSS drivers far outweighs any competitive advantage given. If you can build a functional driver through reverse-engineering alone, you aren't going to be giving away any secrets by just opening up the driver. Seriously... anyone who intentionally renders their hardware worthless for that reason can go sodomize themselves with a hot iron poker.

Well said my friend!

[sylware]

On this very forum, a nvidia driver coder told us that they account mostly GNU/Linux on the professional market.
Meaning that as soon as the AMD driver gets honorable results, it is very probable that a significant part of that market will start switching to AMD. And AMD GPUs have been the best for nearly 2 years and even if nvidia manages to take again the palm, it will never be significantly better to justify sticking to their GPL-violation-but-tolerated/hate-generator blob.

[Ex-Cyber]

It's ugly and incomplete but it does support all NVIDIA GPUs and it IS open source.

I guess that depends on your definition of "open source". The Open Source Definition explicitly forbids obfuscated code (including e.g. only supplying preprocessor output, which some of the nv obfuscation resembles), and the GPL implicitly forbids it via its definition of "source code". To be clear, I don't think this is actually a licensing problem in practice since X clients aren't generally considered to be linking with the server, but it strains some assumptions that are typical for "open source".

[sylware]

I guess that depends on your definition of "open source". The Open Source Definition explicitly forbids obfuscated code (including e.g. only supplying preprocessor output, which some of the nv obfuscation resembles), and the GPL implicitly forbids it via its definition of "source code". To be clear, I don't think this is actually a licensing problem in practice since X clients aren't generally considered to be linking with the server, but it strains some assumptions that are typical for "open source".

Unfortunatly:the driver code is not protected with the GNU GPL. It's MIT/BSD-like which allows to close the code, improve it and make it work better *without* releasing that improved code (cf Darwin/MacOS). For instance the current graphic stack could be "stollen" by apple, secretely improved with hardware secrets provided by GPU manufacturers to be "better" than the stock open source driver. LLVM can suffer easily the same fate for shaders compilation (see the story of the SUN plugin into GCC...). Well there are *very* good reasons to protect code with the GNU GPL: one of them is "we" can legally go for "optimal code". With a MIT/BSD-like license, you have no way to retrieve optimal code... (well... you still can whine and expect multi-million corporations to listen to you), even with the GNU GPL it is hard and often impossible so...

[phred14]

The nouveau drivers enhance the value of nVidia hardware, and presumably will help increase sales. As long as nouveau doesn't cause some sort of legal problems for nVidia, I'd expect them to keep their current approach.

That hands-off approach, by the way, is probably legally the best stance they could possibly take. The moment they get involved, either helping OR hindering, they effectively accept some responsibility for nouveau, if only because in some instance they may have failed to "hinder enough" to satisfy some third party.