NVIDIA Wants To Be A Better Linux Patron

[smitty3268]

You can't discriminately lock out only the nvidia blob just because you don't like nvidia. You'd have to lock all blobs out or risk a lawsuit.

I think Apple has shown pretty conclusively that you can lock out whoever you want, and the 3rd party devs/company can't do anything about it. If that kind of behavior was illegal there's a lot more money to be had going after Apple than Linux. Only MS would have trouble, because they are a monopoly (on the desktop market - they are also locking out developers on their ARM version of windows)

[V!NCENT]

/offtopic
I am interested in ARM for a personal Cloud server + tablet client solution.
This KDE tablet is what I will go with as my first purchase in that arena:
http://makeplaylive.com/ <- the Vivaldi 'Spark' KDE Tablet with open hardware (and reversed engineerd GPU)

The server itself will be probably one of those China open Linux computers. There has to be an open NAS solution possible or I will go with AMD untill that's possible.

The next desktop will be Windows + Linux combination in the form of an AMD Fusion laptop ('ultrabook AMD style'). Let's see if nVidia can get behind Gdev and it will be a regular AMD Fusion+GeForce or Intel Sandy Bitch(if open GPU)+GeForce. But I bet they won't.

/ontopic
I gave my ideas. The discussion is dead. I will conclude that I will buy nVidia hardware if it's opened up, by nVidia, combined with a consotrium based compute standard and Gdev, to serve as a non-DRM compute card. For as long as that is not the case, I will go with AMD. It's up to nVidia.

[XorEaxEax]

How do I argue with somebody who thinks he knows more than me which OS I'm currently running?

Oh I'm not claiming 'I know' which OS you are running, I stated what I 'think' you are running, or rather what I don't 'think' you are running. Of course that had no impact on the discussion at hand so you could just ignore it and discuss the actual points but since you obviously have no arguments you chose to bring up this one-liner as an excuse to flee the discussion. Go ahead, it's not as if I expected you to come up with anything interesting in response anyway.

If you ever get frustrated with nouveau, you can always switch over to the AMD open source driver. I'm sure it's on-par with Catalyst, if not better in every way.

The AMD open source driver won't run on my NVidia cards, obviously.

[johnc]

Snappy comeback... of course it's not as if you had any arguments to begin with.

How do I argue with somebody who thinks he knows more than me which OS I'm currently running?

It's not worth the time.

If you ever get frustrated with nouveau, you can always switch over to the AMD open source driver. I'm sure it's on-par with Catalyst, if not better in every way.

[XorEaxEax]

Snappy comeback... of course it's not as if you had any arguments to begin with.

[johnc]

It gives us the ability to use the hardware we bought on any cpu architecture we want, it gives us a driver which can be fully integrated/debugged/enhanced/updated to reflect kernel driver ABI changes by the kernel devs, and YES, a better Nouveau.


Why not? It's amazing how far the Nouveau devs have gotten with nothing but black box reverse engineering, what makes you think they wouldn't be able to provide a driver(s) of superb quality if they had all the specifications which they now are painstakingly discovering through advanced guesswork? The Nouveau devs are obviously doing a fantastic job, enough even for NVidia to be interested in hiring them. I'm absolutely certain they would be able to provide a driver of rival quality with the full specs at their disposal, it's not as if there's some magic property in the word 'proprietary' even though you seem to think so.


Why wouldn't a great Nouveau driver rivaling NVidia's proprietary driver not benefit those of us (like me) who has bought their hardware? I'm running Nouveau as we speak.


Nothing, they will not lift a finger to help binary blobs as they should not, and the only reason you 'require' binary blobs is because NVidia refuses to open source their drivers or at the very least provide full specs. If you want an operating system for which proprietary drivers are the norm then just use Windows or OSX (and I don't think for a second you are even a Linux user so you are probably using one of them full-time already) or a free operating system which makes it easy for proprietary drivers like FreeBSD, OpenBSD on the other hand has an even harder stance on binary blobs than the Linux devs.

Either that or simply contine to use the proprietary blob NVidia is supplying, not that it exists at all to cater for you as an desktop end user but it's available for you still. Just don't expect NVidia to provide anything directly aimed at the desktop/laptop space for Linux, they won't because they don't give a shit about the Linux end user desktop.

So please johnc, you can stop trying to 'rally the troops' around providing a better situation for binary blob drivers on Linux, it's not going to happen (thankfully!), those few blobs that remain are the last of a dying breed. All I can say is good f***ing riddance, I want to be able to use the hardware I buy in any OS I like (which goes beyond Linux aswell) and not be artificially locked-in on certain platforms due to companies refusing to release the 'secret sauce' necessary to implement drivers which can be ported to any platform.

lol.......

[XorEaxEax]

But again... what is the end game? What does opening up the specs get us? A better nouveau?

It gives us the ability to use the hardware we bought on any cpu architecture we want, it gives us a driver which can be fully integrated/debugged/enhanced/updated to reflect kernel driver ABI changes by the kernel devs, and YES, a better Nouveau.

Ok... a better nouveau is nice... and the kernel devs would be all smiles. But for an overwhelming majority of us NVIDIA end-customers, a better nouveau doesn't help us.

Why not? It's amazing how far the Nouveau devs have gotten with nothing but black box reverse engineering, what makes you think they wouldn't be able to provide a driver(s) of superb quality if they had all the specifications which they now are painstakingly discovering through advanced guesswork? The Nouveau devs are obviously doing a fantastic job, enough even for NVidia to be interested in hiring them. I'm absolutely certain they would be able to provide a driver of rival quality with the full specs at their disposal, it's not as if there's some magic property in the word 'proprietary' even though you seem to think so.

So NVIDIA would spend resources that ultimately would make kernel devs happier, extend an olive branch to the open source community, but in the end won't at all help those of us who forked over hundreds+ to buy their product. How is that fair to us?

Why wouldn't a great Nouveau driver rivaling NVidia's proprietary driver not benefit those of us (like me) who has bought their hardware? I'm running Nouveau as we speak.

What do the kernel devs intend to do to make Linux work better with binary blobs, for those of us who require them?

Nothing, they will not lift a finger to help binary blobs as they should not, and the only reason you 'require' binary blobs is because NVidia refuses to open source their drivers or at the very least provide full specs. If you want an operating system for which proprietary drivers are the norm then just use Windows or OSX (and I don't think for a second you are even a Linux user so you are probably using one of them full-time already) or a free operating system which makes it easy for proprietary drivers like FreeBSD, OpenBSD on the other hand has an even harder stance on binary blobs than the Linux devs.

Either that or simply contine to use the proprietary blob NVidia is supplying, not that it exists at all to cater for you as an desktop end user but it's available for you still. Just don't expect NVidia to provide anything directly aimed at the desktop/laptop space for Linux, they won't because they don't give a shit about the Linux end user desktop.

So please johnc, you can stop trying to 'rally the troops' around providing a better situation for binary blob drivers on Linux, it's not going to happen (thankfully!), those few blobs that remain are the last of a dying breed. All I can say is good f***ing riddance, I want to be able to use the hardware I buy in any OS I like (which goes beyond Linux aswell) and not be artificially locked-in on certain platforms due to companies refusing to release the 'secret sauce' necessary to implement drivers which can be ported to any platform.

[johnc]

BTW, you truncated the post you quoted. It actually finished like this:

Why did you leave out the critical question at the end? Why is it too much to ask for programming specifications?

To be fair, what I quoted was what was posted at the time. The poster afterwards went back and edited his post and added in that question amongst other things. There was no ill will intended on my part.

And to answer your question... I don't know. I don't know that any of us truly knows. I think there's an underlying assumption that releasing such documentation is easy and cheap. It might be... it might not be. It's hard to know for sure without being inside the company and having all of the information. It could be something as simple as their API is so embarrassing that they don't want anyone to know about it. (srs) But I'm assuming that NVIDIA is a rational company motivated by rational reasoning.

It could be that they're working on something now... and don't intend to promise anything publicly until they can follow through. Look how long it took to open source Solaris, even after everyone from the CEO down to the janitor and everyone in between was on board with the idea... it still took forever to navigate through the sea of legal nightmares. In theory, opening up specs would be far less resource-intensive... but you still have all the legal shenanigans that you have to walk through. And lawyers are much more expensive than programmers. (You could probably contract 4 programmers for every 1 lawyer.)

But again... what is the end game? What does opening up the specs get us? A better nouveau? Ok... a better nouveau is nice... and the kernel devs would be all smiles. But for an overwhelming majority of us NVIDIA end-customers, a better nouveau doesn't help us. So NVIDIA would spend resources that ultimately would make kernel devs happier, extend an olive branch to the open source community, but in the end won't at all help those of us who forked over hundreds+ to buy their product. How is that fair to us? What do the kernel devs intend to do to make Linux work better with binary blobs, for those of us who require them?

[hal2k1]

On a 1 GHz single-core CPU?

This will run Netrunner (Ubuntu derivative) comfortably. One of the machines has a 1.5 Ghz CPU. This will run Netrunner very comfortably. Skype even. 1080p video.

Not all users are power users. For mundane office or educational desktop usage, this machine is perfectly adequate. The killer feature is that you can have twenty or thirty for the price of a single Windows 8 machine with equivalent desktop software capability. Kit out the entire class of students for the price of just one Windows 8 machine. No licence hassles whatsoever. No viruses are written to attack this platform ... absolutely none. Extreme low price means throwaway hardware which in turn means no need for expensive maintenance. Ultra low power usage. Unbeatable.

http://www.calligra-suite.org/

http://www.netrunner-os.com/features/

Very nice desktop software. Excellent.

[XorEaxEax]

Then lock out all blobs out of the kernel ... leave only firmware binaries (firmware runs on the card, it isn't part of the kernel).

Well that's where we are right now, the kernel devs have marked large parts of the driver structure symbols as GPL-ONLY meaning that proprietary drivers can't link to them and thus not take advantage of the features they provide (noticeable already two seconds into the boot sequence as NVidia proprietary drivers can't make use of kernel modesetting for example). Also any non-GPL compatible kernel module loaded is automatically flagged as 'tainted' which in practice means that any bugreport encountered containing this flag is generally not investigated as the kernel devs have no interest in trying to debug against a black box.

Here is another plan ... ignore nvidia,

Done, once I upgrade from my current machines, those two with NVidia cards will mark the end of my use of NVidia tech, perhaps also the end of my use of discrete GPU's alltogether given the rapid development of IGP (gpu on cpu) technology.