Edit: I expected downvotes and I'll gladly take them if someone can explain to me how Nvidia is being unreasonable.
Nvidia is being unreasonable in the same general way as Apple is unreasonable with its walled garden and other companies are unreasonable with pollution and poor working conditions. Just because you can get away with something and it makes economic sense does not justify it and does not make it reasonable!
I really don't think we should accept every company's actions just because they make business sense. We, as consumers (and as developers) should promote altruism and good behavior. Being a corporation is no more a license to be a jerk than being a normal person--it's legal in both cases, and we should not condone it in either.
Good communities have their own standards above and beyond the law. In the academic community, plagiarism is not tolerated even if it does not infringe on copyright. The open source community should similarly not tolerate companies and individuals who use the open source software and then refuse to cooperate. And this is all that's needed: cooperation. I don't think anybody even expects Nvidia to write open source drivers; all they want is enough information to not waste time reverse engineering each chip.
In short, Nvidia is unreasonable because it is acting like a jerk. Being legal and making business sense should not justify being a jerk, and we should disparage and avoid companies that act like jerks. So yeah, fuck Nvidia.
If you depend on free software to sell your hardware, it is probably in your interest to see to it that the people that write the software that is required to sell your hardware don't hate you.
Not to disagree with your point (which is right), but It becomes in NVidia's financial interests when they have negative press coverage because someone like Linus says bad things about them.
I hope you understand what I am saying. Linus isn't just talking in an abstract-intellectual environment, his voice is strongly publicised and is here he is using it as an activist, to force NVidia into action. Its all part of the game...
It's not like more than 0.01% potential buyers will read this Linus quote and choose a competing product. Especially if NVidia's product is the best for their use case in the first place.
Less fun edition: I'm sure it is. As it is with Linux based 3D render farms. Do both of those represent a large portion of Nvidia's income or an insignificant one?
I presume it's the latter: there are billions of desktops/laptops but only several thousands of scientific computing / 3D rendering installations. The fact that Nvidia doesn't seem to go out of its way to help people using Linux seems to corroborate that.
And clearly if I had to have worked to some niche fields to know about that use, it's not that widespread.
I seriously doubt Nvidia's bottom line is affected by such professional uses of GPUs.
>And how about a large portion of android phones?
What about them? You think companies like Samsung or Motorola are gonna switch just because Linus said something negative? Or that Android customers are gonna demand a different GPU?
Not to mention that Nvidia is, if not "the only game in town", one of the "counted on the fingers of one hand with some digits cut-off in a tragic saw accident" games in town.
Why should Nvidia devote resources, potentially a significant amount, to make Optimus work on Linux? I do not buy the argument that they should do so out of gratefulness that the Linux based Android OS let's them sell a large number of unrelated chips.
I'm open to being convinced otherwise, but citing a lack of "gratefulness" is not compelling.
That may be true for specific features like Optimus--I can't say. But if you're talking about writing a full 3D driver stack for a modern GPU, you're almost certainly underestimating the effort involved by several orders of magnitude. NVIDIA has hundreds of full-time software engineers working on the various parts of the GPU driver stack. Now imagine doing it without the immense institutional knowledge.
Now try to find a laptop with a discrete Nvidia GPU that doesn't have this Optimus junk and you'll see the problem.
It's not really incongruent. Nvidia has provided driver support for Linux, but that does no good for, say, FreeBSD. This driver support is in lieu of providing specs. They're keeping a lid on things and maintaining control. There are very good hackers who would make open source drivers if they had specs. Arguably less buggy drivers. Since they make sure they have control it's not surprising that they either 1) think this isn't worth their trouble, or 2) have actively decided they don't want this to happen, for whatever business reason.
There was a stink a while back where Adaptec had been providing drivers for FreeBSD, but then stopped. People were stuck either not upgrading their boxes (bad!) or moving off their RAID setup (painful, expensive!). It's ugly. With specs, there is not this issue.
The (pervasive) idea that vendors providing drivers for your system of choice, for the time being is being open source friendly is something the vendor promotes, but it's not actually friendly.
It's worth noting that just last night they released a stable driver that (finally) supports RandR 1.2. From what I understand this was also a significant change and it may bode well for future standardization improvements like KMS (moves multi-monitor management away from the proprietary TwinView driver components into standard xrandr, for instance, so that is complexity the KMS version will not have to handle).
Fingers crossed that they release this imaginary driver soon. :)
The XRandR support in the driver is not yet stable [1]. The 302.xx line is currently in beta, but will support both XRandR 1.2 & 1.3 at the same time. That part also works flawlessly, however, your mileage may vary in regards to suspend/hibernate, since both I and multiple others on the nvidia forums experience serious problems with resume from either, i.e. the computer hard locks (and doesn't even respond to ssh etc.), which requires you to powercycle it. So while the XRandR is a huge step forward, it is not without problems yet.
[1]: ftp://download.nvidia.com/XFree86/Linux-x86_64/latest.txt shows you the latest stable driver, which is 295.59
ArchLinux has already sent out packages for this release in its stable update channel: https://projects.archlinux.org/svntogit/packages.git/commit/...
(As an aside, it still hardlocks my computer...but according to their forums, they "are working on a fix" for the soon-to-come 304.*)
I guess he's talking about pushing out some new driver support.
support
Using a fragment identifier to demarc the time offset works better for me:
Her question actually starts a minute earlier than the link:
http://www.youtube.com/watch?v=MShbP3OpASA&t=48m14s
PS - Bravo Linus. This is issue is a real PITA, and a bit incongruent considering the historically awesome driver support Nvidia has provided for Linux.