KDE's Nate Graham On X11 Being A Bad Platform & The Wayland Future

[mrg666]

I don't even notice I use Wayland anymore, nor I miss any feature from X11. So glad that I don't need that unmanageably aged pile of code for anything. I don't have any fond memories of hopelessly editing X11 configuration files to achieve a useable and clear scaling on hidpi monitors, trying to adjust fonts and cursor sizes, trying to use never working features such as remote GPU acceleration, turning it off for security on servers, etc, etc.

[Volta]

I don't even notice I use Wayland anymore, nor I miss any feature from X11. So glad that I don't need that unmanageably aged pile of code for anything. I don't have any fond memories of hopelessly editing X11 configuration files to achieve a useable and clear scaling on hidpi monitors, trying to adjust fonts and cursor sizes, trying to use never working features such as remote GPU acceleration, turning it off for security on servers, etc, etc.

Yeah, I remember generating some xorg.conf madness to make nvidia blob working. Oh, and such insecure, full of holes pile of mess ran as root for most of the time. Thankfully it's a horror of the past. EDIT: V1tol reminded me about another brokeness of Xorg. To have good performance desktop compositions had to be disabled for games.

[Steffo]

I don't remember Linux developers defending systemd, pulseaudio/pipewire or devtmpfs. They got the transition right. They extended and improved a lot on what was before them.

You are saying systemd and pulseaudio weren't controversal? In what universe?!

Wayland initially did not extend, it was designed with a lot less stuff which is required for the average desktop user.

Wayland wasn't designed for the desktop. It can be used for the desktop, but also in the embedded world. Wayland is transparent.

And to this date it doesn't have a killer feature which is required by lots of users. HDR is still not there.

KDE Plasma 6 has an implementation.

Per monitor scaling? Yeah, but at what cost?
DPI scaling in X11 is miles better and versatile than fractional scaling in Wayland.

Why?

And it doesn't help that each Wayland compositor is basically a whole X.org server. That's just pure insanity and probably the reason the progress in Wayland adoption has been so slow.

This is not true. The Wayland compositor does not everything. The work of the X.org server is split up into compositor, graphical tool kits (Qt, GTK), Portal and Pipewire.
This is much more modular and secure. I mean, you can easily implement a key logger in python for X with a few lines of code. In Wayland? Not that easiy.

I'd say scrap this crap. Sorry. Look at how Microsoft did it with Windows Vista where they completely replaced the entire graphics stack. Maybe X11 did not have to be fully replaced but it needed to be rethought and readjusted, not killed off.

Windows has the most inconsistent and crappy UI I have ever seen!

[kpedersen]

Yeah, I remember generating some xorg.conf madness to make nvidia blob working. Oh, and such insecure, full of holes pile of mess ran as root for most of the time. Thankfully it's a horror of the past.

What solved this was actually the libdrm / modesetting drivers (which first appeared on Xorg and is identical on Wayland compositors). This has nothing to do with the higher level display systems.

If individual wayland compositors had to come up with a way to support usermode drivers it would have been much, much worse.

[V1tol]

As a user I am happy to get rid of X11 for 2 reasons - tearing (modesetting driver received TearFree only a year ago!) and need to disable compositor for fullscreen games to maintain performance. On Wayland things just worked even 2 years ago. I can game and alt-tab to check something without looking at ugly tearing windows without shadows because compositor is disabled. And I don't need to force TearFree which causes different stutters in games. Wayland restrictions, force use of portals? I only welcome that because I, as a user, want to control which apps have access to my screen, keyboard or files.

[dlitz]

I don't remember Linux developers defending systemd, pulseaudio/pipewire or devfs.

hahahahaha. Developers are still arguing about systemd. There have been entire distro schisms over it. Life goes on.

devfs, has been since Linux 2.6. I dunno why you'd mention it. Perhaps you don't remember discussions about udev on Phoronix about it because Phoronix didn't exist yet? There was plenty of discussion about devfs on lkml, Slashdot, and elsewhere. People also hated pulseaudio for a long time due to latency issues.

Maybe just maybe someone has to admit they royally fucked up with what was necessary to fully replace X.org and make the replacement enticing for the user.

Users complaining doesn't mean Wayland has a user adoption problem. People are complaining now because they're running it, much like what happened with pulseaudio when distros started shipping it by default. Lots of people (myself included) uninstalled it, at first, until their corner of the ecosystem caught up.

As with PulseAudio (and ALSA before that), users will run it when distros ship it and their apps work with it. Eventually, there will be a tipping point where more developers care about their stuff working in Wayland than in native X11.

Look at how Microsoft did it with Windows Vista where they completely replaced the entire graphics stack.

Yeah, they included a compatibility layer for the old interface (GDI). Same as what XWayland represents. The compositor/windowing system still had to be rewritten to use the new API, and Microsoft didn't have to get buy-in from multiple DE projects or to ship half a dozen different-yet-compatible compositors in order to be considered viable...

These things take a while and it's only been about 15 years since Wayland was first announced. For comparison, Python 3 was announced about 17 years ago, and didn't change things nearly as fundamentally as Wayland does.

At this point, Wayland basically works, and it just needs to be brought up to feature parity with Xorg.

Maybe X11 did not have to be fully replaced but it needed to be rethought and readjusted, not killed off.

That's what Xorg is. You're welcome to join the Xorg project and to "rethink and readjust" it yourself, if you think that's so much more viable than doing feature development on Wayland. Protocol specifications and running code speak louder than complaints, and if you think that's so easy, now's the time to get involved and show us all what we've been missing, while the code is still being actively developed.

[pabloski]

What solved this was actually the libdrm / modesetting drivers (which first appeared on Xorg and is identical on Wayland compositors). This has nothing to do with the higher level display systems.

This is exactly why X11 is crap. The Linux community was forced to implement the low level stuff into the kernel and strip it from Xorg!!! This is what some has been saying in past comments, that Xorg isn't what it used to be at all. It is a massively stripped down version of the original X11/Xorg.

[timofonic]

This is exactly why X11 is crap. The Linux community was forced to implement the low level stuff into the kernel and strip it from Xorg!!! This is what some has been saying in past comments, that Xorg isn't what it used to be at all. It is a massively stripped down version of the original X11/Xorg.

And? That's when Xorg got massively better! Before that, it was even worse

[jeisom]

The biggest problem that has been holding things back is the glacier pace of protocols being added that would be considered required. Fractional scaling was only agreed upon in the last year or so. A copy/paste protocol that works on Wayland wasn’t added until the last couple years. This doesn’t even get to the fragmentation right now that is probably slowing things down. Sure nvidia dug their heals a while, but what was available was not a usable for most users even on “supported” gpus until the last couple years.

Plasma 6 beta 1 was looking really good, but then beta 2 came out and it seems like a step backwards, at least on Arch. KDE needs to implement a “refresh plasma” setting feature.

Wayland is the future and I look forward to using it more this year. That said it has taken over a decade of development, protocol and compositors, to get to this point.

[WileEPyote]

hahahahaha. Developers are still arguing about systemd. There have been entire distro schisms over it. Life goes on.

devfs, has been since Linux 2.6. I dunno why you'd mention it. Perhaps you don't remember discussions about udev on Phoronix about it because Phoronix didn't exist yet? There was plenty of discussion about devfs on lkml, Slashdot, and elsewhere. People also hated pulseaudio for a long time due to latency issues.



Users complaining doesn't mean Wayland has a user adoption problem. People are complaining now because they're running it, much like what happened with pulseaudio when distros started shipping it by default. Lots of people (myself included) uninstalled it, at first, until their corner of the ecosystem caught up.

As with PulseAudio (and ALSA before that), users will run it when distros ship it and their apps work with it. Eventually, there will be a tipping point where more developers care about their stuff working in Wayland than in native X11.



Yeah, they included a compatibility layer for the old interface (GDI). Same as what XWayland represents. The compositor/windowing system still had to be rewritten to use the new API, and Microsoft didn't have to get buy-in from multiple DE projects or to ship half a dozen different-yet-compatible compositors in order to be considered viable...

These things take a while and it's only been about 15 years since Wayland was first announced. For comparison, Python 3 was announced about 17 years ago, and didn't change things nearly as fundamentally as Wayland does.

At this point, Wayland basically works, and it just needs to be brought up to feature parity with Xorg.



That's what Xorg is. You're welcome to join the Xorg project and to "rethink and readjust" it yourself, if you think that's so much more viable than doing feature development on Wayland. Protocol specifications and running code speak louder than complaints, and if you think that's so easy, now's the time to get involved and show us all what we've been missing, while the code is still being actively developed.

Agreed.

Speaking in generalizations for this post.

I remember all of these things too. The transition from ALSA to Pulse was a tough one for me with my audio interfaces. Once Pulse came to performance parity for me, I stopped replacing it with ALSA. I didn't cry about it when it shipped as the standard in a distro, I just replaced it and got on with my life.

It's the same for any new protocol. I'm personally in a Wayland session typing this. Why? Because it works, and works well for me. I still have x11 installed in case I need it, but I honestly can't remember the last time I logged into an X11 session. If it doesn't work for you, just don't use it.

All this arguing over these protocols is stupid. Wayland is here to stay until the next big thing comes along. This is how development always was, and always will be. The beauty of Linux is that you don't have to use what you don't want to use. If you don't like Wayland, cool, don't use it. We haven't gotten anywhere near the point where you're forced to do so. Just stop arguing over it and crying about how much you disagree with it.

Instead of arguing and crying about things, how about try contributing something useful? If you can't code, it can as simple as filing bug reports so they can get things ironed out and make it run better for everyone.