NVIDIA EGLStreams Support Merged Into KWin For KDE Plasma 5.16

[carewolf]

"Wayland" isn't the compositor, Kwin is. And Kwin can extend the Wayland spec all it wants to implement the features people need to actually use it.

The problem is that simultaneously there are at least 4 disparate Wayland display server implementations in development, all with different extensions and features, and all reimplementing the same things over and over between them.

With X there was just one X server and all the window manager clients who did generally end up having to do a substantial portion of what was once X's job but I still think it is a failing of Wayland in general to not have the One True Display Server to make this transition less time consuming.

No, Wayland could have plenty of display servers. The problem is that too much is left unspecified and up to the display server, making huge parts of "Wayland as an X11 replacement" unspecified, and thus display server specific.

[Azrael5]

You're talking like this is a new thing with Wayland. Remember how many compositors there were for X? How many window managers there were? Remember xlib and xcb, the two different libraries for talking with X? Remember how you sometimes had to use both? And let's not forget how we had X drivers as well as mesa drivers and kernel drivers. The only reason X seems so smooth by comparison is because Wayland is reimplementing the results of 30 years of organic development without the cruft. And this time around, it's actually being *designed* instead of bolted on in an undocumented fashion. There's also committees and collaboration across compositors and toolkits this time. But at least everyone's aware of the full scope of the project from the beginning.



What, like Weston? Even if Wayland decided to mandate that as the "one true display server", people would make their own. KDE, Gnome and Enlightenment have very different ideas about code quality, programming languages, visual styles, feature sets, etc. They don't even agree on window decoration strategies. Their users already kvetch massively if they have to load the libraries from another toolkit to display an application. They don't even like having those libraries taking up disk space. That's why both toolkits have had "toolkit-native" implementations of the same programs as opposed to just agreeing to let KDE do one app and Gnome do another.

Let's be real. The problem about people collaborating on one implementation as opposed to multiple competing ones is a community problem, not a Wayland problem. They'll complain until they get an implementation they like with the libraries and languages they like, and only then will they tell other people that there's enough implementations. Not only that, they'll want to combine an unholy, fragile mix of applications that they're used to instead of having low-overhead, configurable defaults. They'll also complain about how you configure things. And no one will agree on which app should be the one true implementation. Regardless of whether it's the first one, the last one, or the most maintained one. Yup, WM/DE user wars are pretty much systemd wars with wobbly windows. Complete with the ignorant trolls.


Wayland stayed away from implementing everything for compositors on purpose. Most of the wayland devs were X devs, and they remember a lot of the bigger issues they dealt with.

X was brittle - hard to extend and hard to maintain. It had a lot of useless stuff baked in that seemed really essential at the time it was made. Wayland was designed to mandate as little as possible, and to add extensions on piece by piece until everything was finished. And there are committees. Lots of them. It's going to take a while, but it's also going to be way better. Good news is that it's already fairly usable. Bad news is that the community is a dumpster fire, as usual. But it's our dumpster fire.

Instead of making the same different project every team should collaborate in only one project. Discussion and choices, then verification of the result and feedback, subsequently study of alternatives in efficiency and effectiveness. I think that teams needs managers or management to make their projects.

[Terrablit]

Instead of making the same different project every team should collaborate in only one project. Discussion and choices, then verification of the result and feedback, subsequently study of alternatives in efficiency and effectiveness. I think that teams needs managers or management to make their projects.

You vastly overestimate our ability to agree on implementation details, as well as the general Linux community's ability to not trash on anything popular that makes changes to The Great Unix Dogma. The hipsters that think computer science peaked in the 1970s tend to be resistant to vertical organizational hierarchies, corporate-funded contributions, and configuration DSLs that would require them to learn something other than shell script.

I'm not saying that you're wrong, but I do think that it's easier to propose than implement.

[oliver+]

You're talking like this is a new thing with Wayland. Remember how many compositors there were for X? How many window managers there were? Remember xlib and xcb, the two different libraries for talking with X? Remember how you sometimes had to use both? And let's not forget how we had X drivers as well as mesa drivers and kernel drivers. The only reason X seems so smooth by comparison is because Wayland is reimplementing the results of 30 years of organic development without the cruft. And this time around, it's actually being *designed* instead of bolted on in an undocumented fashion. There's also committees and collaboration across compositors and toolkits this time. But at least everyone's aware of the full scope of the project from the beginning.



What, like Weston? Even if Wayland decided to mandate that as the "one true display server", people would make their own. KDE, Gnome and Enlightenment have very different ideas about code quality, programming languages, visual styles, feature sets, etc. They don't even agree on window decoration strategies. Their users already kvetch massively if they have to load the libraries from another toolkit to display an application. They don't even like having those libraries taking up disk space. That's why both toolkits have had "toolkit-native" implementations of the same programs as opposed to just agreeing to let KDE do one app and Gnome do another.

Let's be real. The problem about people collaborating on one implementation as opposed to multiple competing ones is a community problem, not a Wayland problem. They'll complain until they get an implementation they like with the libraries and languages they like, and only then will they tell other people that there's enough implementations. Not only that, they'll want to combine an unholy, fragile mix of applications that they're used to instead of having low-overhead, configurable defaults. They'll also complain about how you configure things. And no one will agree on which app should be the one true implementation. Regardless of whether it's the first one, the last one, or the most maintained one. Yup, WM/DE user wars are pretty much systemd wars with wobbly windows. Complete with the ignorant trolls.


Wayland stayed away from implementing everything for compositors on purpose. Most of the wayland devs were X devs, and they remember a lot of the bigger issues they dealt with.

X was brittle - hard to extend and hard to maintain. It had a lot of useless stuff baked in that seemed really essential at the time it was made. Wayland was designed to mandate as little as possible, and to add extensions on piece by piece until everything was finished. And there are committees. Lots of them. It's going to take a while, but it's also going to be way better. Good news is that it's already fairly usable. Bad news is that the community is a dumpster fire, as usual. But it's our dumpster fire.

Thank you

[Azrael5]

You vastly overestimate our ability to agree on implementation details, as well as the general Linux community's ability to not trash on anything popular that makes changes to The Great Unix Dogma. The hipsters that think computer science peaked in the 1970s tend to be resistant to vertical organizational hierarchies, corporate-funded contributions, and configuration DSLs that would require them to learn something other than shell script.

I'm not saying that you're wrong, but I do think that it's easier to propose than implement.

Indeed, vertical organizations currently are unreliable, are obsolete. Nowadays horizontal management solution are implemented so to highlight the activities and the final result in efficiency and effectiveness of the processes useful for the final user. One of the best management system method is the time-driven ABC on how to set the organization and the ABM so called activity based management based on the distinction among value added activities and non-value added activities by the PVA so called Process Value Analysis. The useless activities, the redundancies are identified and deleted from the processes in order to save time, so to improve efficiency, to improve the result for the final user and improve the ability and the return of the processes where the activities operate. In this kind of business model, the only residual vertical activity are control, planning and programmation. The organization is horizontal and based on the supply chain of activities able to produce internal and external results where everyone and everything have their own competence. Linux development business misses in organization producing 20% of result by 100% of resources. This condition means also that every developer wastes 80% of time of his life although its contribution is free... however the 80% of free time would be used to improve linux in profitable way (I don't allude to money but in better result): too much energy loss. Manager understand all of this critical limit and if he is not interested only in money, he represents the resource able to make an organization able to reach the maximum result by the optimal utilization of resources: 100% of result by 100% of the resources.

[Vistaus]

KWin on Wayland is still not ready, due to:

https://bugs.kde.org/show_bug.cgi?id=387313

https://phabricator.kde.org/T10530

Same flickering bug on X11 with 5.15.4 (didn't happen on 5.15.3), so not a Wayland-related bug per se.

[Jabberwocky]

You're talking like this is a new thing with Wayland...

Thanks for taking the time, I feel many people needed to read this. I have used xlib and xcb in the past, but know very little about Wayland besides from what I read on Phoronix. Below are some examples why people are afraid of things that Nvidia are part of.

TL;DR Nvidia has questionable business practices, people are afraid that they will abuse Wayland and/or the implementations of it.

Everyone makes honest mistakes. Some plan to mislead people and just hope that they don't realize. Nvidia have been caught with large scale production of hardware that do not meet the specifications, cheating in game benchmarks, and getting game studios to make titles perform worse on competitors hardware. When exposed they do not admit their faults. The Geforce Partner Program has still not been made public, Nvidia just says: "Rather than battling misinformation, we have decided to cancel the program". It does not take a genius to see that they are hiding things.

References:
2003 - https://www.geek.com/games/futuremar...chmark-553361/ (ATI was also cheating at this stage, image quality instead of obj culling hacks)
2011 - https://techreport.com/review/21404/...a-good-thing/2 and https://www.youtube.com/watch?v=IYL07c74Jr4
2013 - Nvidia Optimus (Close to this thread's topic, most of us knows what happened)
2013 - https://www.extremetech.com/extreme/...sers-and-amd/2
2018 - https://www.pcworld.com/article/3269...r-program.html
2018 - https://www.gamersnexus.net/game-ben...ameworks-tests

In my experience Nvidia has not changed much over the past ~decade besides becoming more cunning. With open source software these types of things are more apparent at least.

[edoantonioco]

still waiting for wayland kde experience to be as good as xorg experience, which is not since I have tried every release

[DanL]

In my experience Nvidia has not changed much over the past ~decade besides becoming more cunning. With open source software these types of things are more apparent at least.

Yeah, and they change their pattern of behavior depending on who is making the decision (like with the nouveau firmware). That is the problem with too much bureaucracy.

[Raka555]

The problem is Wayland caused a mess.
It is way too simple to replace X.

It basically forced all the X11 toolkits to evolved into tools to create wayland "servers" and do it however they felt like.
(CSD comes to mind...)

The result is that these toolkits has become so bloated (even under X).