News for hackers. Stuff that matters.
Re:Long live Xorg, I mean Xlibre by bsder on Tuesday June 17, 2025 @05:42PM (#44304266)
> If you like Xorg, use Xorg. If you like Wayland, use Wayland. If you're not happy about an issue, contribute to it.

The problem is that RedHat nee IBM are attempting to force everybody onto Wayland by dropping X11 support. They already tried once and the outcry was so huge that they had to back off saying they would try again next version.

This is kind of a rock and a hard place. The Wayland developers don't want to support X11 but neither does anybody else. Wayland is fundamentally broken in many ways down at the architectural level, but the sunk cost fallacy keeps them working on it.

Everybody forgets that Wayland predates Vulkan. A "real" replacement for X11/Wayland probably needs to restart from "Vulkan support is the base layer" and build up from there.

Re:Long live Xorg, I mean Xlibre by em-bee on Tuesday June 17, 2025 @06:33PM (#44304706)
Wayland predates Vulkan

and how is that waylands fault? what if vulkan gets replaced? should a new replacement be written from scratch every time that happens?

Re:Long live Xorg, I mean Xlibre by bsder on Tuesday June 17, 2025 @10:04PM (#44306021)
> and how is that waylands fault?

It is not. However, Wayland had to make certain decisions up front that later turned out to be significantly suboptimal in the context of a graphics world that converged to DX12/Vulkan/Metal and all look roughly the same from an abstraction level.

Re:Long live Xorg, I mean Xlibre by yjftsjthsd-h on Wednesday June 18, 2025 @12:09AM (#44306584)
> Wayland is fundamentally broken in many ways down at the architectural level, but the sunk cost fallacy keeps them working on it.

> Everybody forgets that Wayland predates Vulkan. A "real" replacement for X11/Wayland probably needs to restart from "Vulkan support is the base layer" and build up from there.

I wonder if you could point me in the right direction to learn more about this?

Re:Long live Xorg, I mean Xlibre by bsder on Wednesday June 18, 2025 @02:52AM (#44307285)
Erm, what do you want to learn? Wayland dates to 2008--Vulkan dates to 2016. That's an EON in tech.

Wayland chose "we will control the compositing to avoid all tearing". This is suboptimal nowadays with monitors with multiple refresh rates some of which are on completely different graphics cards. It's better to let an application have a Vulkan (or equivalent) surface, draw to it, query, and choose its own linkage to the refresh rate. Also, nobody other than Wayland devs give one iota of damn about performance or tearing during resizes (see: macOS and Windows). There's also the issue of how much resource Wayland compositors have to overallocate in order to be able to composite everything. I have applications that regularly crash because Wayland chewed up too much VRAM and the application couldn't get sufficient VRAM resources.

It's better to let something like Zink handle OpenGL rather than try to provide an OpenGL interface. It's better to give something a surface and let Vulkan handle video rendering onto the surface (the application understands whether it needs to do something like sync to audio or whether it is just an animation). etc. This is even more especially true in the world of massive multicore/multithread.

Then there are the architectural decision that range from the silly to the flat out wrong. Fractional scaling is handled by doubling and shrinking--that's a huge waste of processing. Gnome idiocy around client-side decorations which are just a disaster because it forces everybody to link to C/C++ libraries even if they aren't using C/C++.

Security stuff that has been a disaster: Stopping or arbitrarily jiggling the refresh rate to prevent "application fingerprinting" is dumb and just pisses everybody off. Pointer and focus issues because you might figure out some way to intercept people typing. Screenshot/screencast is an absolute PITA because "security". Input Method and accessibility that threw out a ton of working stuff from X11 that still hasn't been replaced.

I mean, the Wayland developers are not wrong that these things are security issues ... but you can't let them continue to persist for more than a decade and then wonder why people think your project sucks. If you don't give them something, then don't complain when they go back to the old, shitty, insecure thing that works.

As you can see, there are a LOT of corner cases to be covered. This is the primary problem. The video subsystem is a lot of fiddly work that very few people are qualified to do. Look at how much work Asahi Lina and Alyssa Rosenzweig had to do to make graphics work on Linux on Apple hardware. Now, multiply that out and you can see just what a mountain this stuff is.

Be aware that in the discussion above, Vulkan can be substituted with Direct12 or Metal which are effectively the exact same graphics API.

Which is actually probably the end result of all of this. We're probably going to wind up with some version of Win32/DX12 APIs on Linux thanks to Steam as that would give you portability betweeen Linux and the biggest OS platform. And, to be fair, most users would probably be happier...

Re:Long live Xorg, I mean Xlibre by palata on Wednesday June 18, 2025 @03:00AM (#44307327)
Those are interesting points. Would be great if someone started such a better alternative.
Re:Long live Xorg, I mean Xlibre by solid_fuel on Wednesday June 18, 2025 @05:14PM (#44313331)
> Vulkan can be substituted with Direct12 or Metal which are effectively the exact same graphics API

Just for clarity, have they actually converged to the point where the APIs are that similar? Or do you just mean they fill the same role?

Re:Long live Xorg, I mean Xlibre by bsder on Wednesday June 18, 2025 @05:59PM (#44313626)
They have converged to the point that the APIs are extremely similar. To the point that you can add a translation layer from Vulkan to Metal (MoltenVK) and lose very little performance, for example.

The concurrency primitives are very similar. The way they all manage memory is a little different, but not horribly so.

And now that DirectX and Vulkan are both supporting SPIR-V as the shader intermediate format, things really have converged.

Re:Long live Xorg, I mean Xlibre by NM64 on Tuesday June 24, 2025 @02:59AM (#44363534)
Don't forget that AMD's Mantle API is the common ancestor for both Vulkan and D3D12.
Re:Long live Xorg, I mean Xlibre by NM64 on Tuesday June 24, 2025 @02:58AM (#44363530)
I'm not OP, but it's worth noting that at least Vulkan and D3D12 have a common ancestor in the form of AMD's Mantle API.
Re:Long live Xorg, I mean Xlibre by kloop on Friday June 20, 2025 @12:33AM (#44324778)
> The Wayland developers don't want to support X11 but neither does anybody else.

You can't say that in a thread about the main maintainer forking it because he wants to maintain it against the wishes of the original maintainers.

This post is literally about someone wanting to maintain it.

Re:Long live Xorg, I mean Xlibre by creatonez on Sunday June 29, 2025 @11:01PM (#44418853)
The maintainer of Xlibre was not the "main maintainer" of Xorg. It is all smoke and mirrors, a fake-it-till-you-make-it strategy to be able to prematurely say "I am a X developer" without having put in the work. Their commits only go back to 2022 and they are all miniscule changes that haven't fixed any of Xorg's architectural warts.