KWin Maintainer On KDE Wayland Remains Uninterested In NVIDIA's Driver

[RealNC]

Why do you expect to do Nvidia's work? He said that Nvidia is free to do it. KDE's budget is tiny compared to Nvidia's multi billion profit . What's stopping them if they cared?

Ohh.. yeah . You have an Nvidia card and you don't want to have second thoughts of your purchase.

There's no alternative for nvidia gpus for my use case right now.

However, my issue with Martin isn't that he's not adding support for it. That's totally fine. What ticked me off is his attitude of not accepting patches unless they come directly from nvidia. Open source at its best, eh? Coming up with BS terms of accepting patches.

That's the issue I have with Martin. BS attitude and BS reasoning.

"I won't implement this." -- Completely fine. No one should be forced to work on something they don't want to when they're a volunteer.

"I won't allow other people to implement this." -- Not fine. Compete BS, anti-open-source and damaging to KDE.

[carewolf]

Qt already supports Eglstreams, so at least KDE applications should work with a hypothetical compatible compositor.

[nils_]

However, my issue with Martin isn't that he's not adding support for it. That's totally fine. What ticked me off is his attitude of not accepting patches unless they come directly from nvidia. Open source at its best, eh? Coming up with BS terms of accepting patches.

If you think about long-term maintenance it doesn't sound so unreasonable. And it's not like we're discussing something that's likely to happen.

What I really have a problem with is people who own Nvidia hardware complaining about Martin not doing free work for them while their beef should really be with Nvidia whom they paid a tidy sum of money for hardware that isn't future-proof (basically part of their business model).

[TheBlackCat]

However, my issue with Martin isn't that he's not adding support for it. That's totally fine. What ticked me off is his attitude of not accepting patches unless they come directly from nvidia. Open source at its best, eh? Coming up with BS terms of accepting patches.

After the number of times kwin has been burned by someone implementing a big, hard-to-maintain feature and then disappearing, this sort of position is pretty understandable.

[oiaohm]

However, my issue with Martin isn't that he's not adding support for it. That's totally fine. What ticked me off is his attitude of not accepting patches unless they come directly from nvidia. Open source at its best, eh? Coming up with BS terms of accepting patches.

Lets break this down into a few points.
1) Only video card maker using egl stream is Nvidia.
2) Only party that can really maintain a full build farm for testing is Nvidia.
3) Nvidia has not stopped messing with how wayland and egl streams interact.
4) Nvidia egl streams has a lot of what in hell.

The second point is kind of critical Nvidia could make a new copy of a non longer mass produced card for their testing system really no one else can.

https://github.com/NVIDIA/egl-wayland/blob/master/wayland-eglstream/wayland-eglstream.xml

https://www.khronos.org/registry/EGL/extensions/NV/EGL_NV_stream_socket.txt

So how are you going to test this when it has a ipv4 usage in a undocumented format.

So like or not the patches need to be signed off by Nvidia also what is up with this undocumented network streaming format that could possible be used to spy on end users.

There is no way I would want to sign off on something that could be a huge security hole either. I would want the designer of it be blameable if something is wrong.

[Guest]

I don't have a problem with that. Nor am I worried about Nvidia's stance. I'm pretty sure if and when Wayland becomes the de facto standard, they'll have working support for it.



They wrote the first video driver for Linux that didn't sucked. And they did it before Mesa. Linux would be ruling the desktop world by now if everyone would be leeching off it like Nvidia

True. The main reason Linux kernel is getting contributions is because it is a free product that companies can base commercial products on (Android, cloudlinux, RHEL, etc...). Linux would die if no one could "leech" of it.

[bug77]

There's no alternative for nvidia gpus for my use case right now.

However, my issue with Martin isn't that he's not adding support for it. That's totally fine. What ticked me off is his attitude of not accepting patches unless they come directly from nvidia. Open source at its best, eh? Coming up with BS terms of accepting patches.

That's the issue I have with Martin. BS attitude and BS reasoning.

"I won't implement this." -- Completely fine. No one should be forced to work on something they don't want to when they're a volunteer.

"I won't allow other people to implement this." -- Not fine. Compete BS, anti-open-source and damaging to KDE.

Tbh his reasoning is more along the lines of "Nvidia started this, they should end it" - he doesn't want people wasting time on something that's entirely on Nvidia. So nobody tells him what he should be working on, but he has some guidelines for what other should do with their time - a little ironic if you ask me.

Anyway the whole discussion is moot: no third party will submit those patches, because Nvidia has been working the whole year and they're still nowhere near ready. I doubt a bunch of open source enthusiasts can do better. So I suggest we just close this subject for good.

[Ansla]

Open source at its best, eh? Coming up with BS terms of accepting patches.
"I won't allow other people to implement this." -- Not fine. Compete BS, anti-open-source and damaging to KDE.

He didn't actually say that he will not accept a patch from somebody other than nVidia, just that he doesn't what anybody else to waste their time doing it. Really, do implement it and send the patches, I'm sure they will be accepted if they are technically good.

[ResponseWriter]

It seems progress is being made but very slowly: https://blog.martin-graesslin.com/bl...#comment-74234

I'm with Martin on this. The information is available for people to make a choice whether to buy Nvidia, and forgo acceleration on XWayland, or something else. The lack of acceleration in XWayland is Nvidia's problem because they write the driver that doesn't support these established interfaces. That's up to Nvidia to fix, not people like Martin and others who have limited time and resources to maintain these other interfaces.

Xorg is going to be supported for years to come. Your Nvidia graphics cards are not going to suddenly stop working just because everyone else moved to a newer display server.

[oiaohm]

It seems progress is being made but very slowly: https://blog.martin-graesslin.com/bl...#comment-74234

I'm with Martin on this. The information is available for people to make a choice whether to buy Nvidia, and forgo acceleration on XWayland, or something else. The lack of acceleration in XWayland is Nvidia's problem because they write the driver that doesn't support these established interfaces. That's up to Nvidia to fix, not people like Martin and others who have limited time and resources to maintain these other interfaces.

Xorg is going to be supported for years to come. Your Nvidia graphics cards are not going to suddenly stop working just because everyone else moved to a newer display server.

Thanks for reminding me. https://www.phoronix.com/scan.php?pa...Device-Mem-API Unix Device Memory Allocation Library this is a proposed idea by Nvidia that could see egl stream stuff deprecated. This is a big problem with stuff from one vendor as they upstream it can cease to exist due to push back from other vendors as well. Now this new allocation method from Nvidia could see standard GDM interface work with Nvidia hardware.