the classic 'perl is dead' argument. things get feature rich and basically stable. Especially for dull libraries. CPAN has got an option for everything, new language X's package collection? Maybe, maybe not.

Nvidia doesn't seem to care :(

Why should they? ML or gaming on Linux doesn't give a shit about Wayland.

Because X has inherent design flaws. E.g. security related.

Almost every commonly deployed system does.

They've had feature parity forever. Every app ever released hasn't been updated yet, however.

Wayland's accessibility story is... Crap. In the name of security they killed most of the ways you'd expect to be able to pass on information to users who fall anywhere outside the norm.

For example, you can't fetch the mouse position, you can't fire mouse events, you can't get the window z-index. These are all pretty vital for accessibility.

Whilst there are efforts to get around this terrible-ness, you can expect most accessibility tools to segfault or misbehave.

There are ways to do all these things but they are compositor agnostic - which frankly makes sense. I don’t want a random app running on my desktop querying these informations in a hacky way, while the whole of this information is stored nicely in memory by the compositor. They just simply have to provide an API for that - and that’s the beauty of wayland - they let the compositor implement any number of custom apis and the app can query which ones are supported by the compositor and vice versa.

> There are ways to do all these things but they are compositor agnostic

Maybe. But GNOME have found it difficult [0] [1], and the most basic of mouse control required a truly massive patch to KDE [2]. Everyone else just shrugs and says to use XWayland. Meaning there is no consistency on how to make basic accessibility work.

It isn't clear to me why reporting the mouse's location belongs in the compositor. You may not be rendering to a screen at all.

The end result being... Accessibility under Wayland is crap. Accessibility under X11 mostly works.

[0] https://gitlab.gnome.org/GNOME/mutter/-/issues/9

[1] https://bugzilla.gnome.org/show_bug.cgi?id=709999

[2] https://phabricator.kde.org/D11468

Sounds like the beauty of a lump of rock - any sculpture you can imagine could potentially "just simply" be carved from it.

To replace X11, then, we would need something more than Wayland: a design (as a spec or as a concrete component) for how all these parts that were left out are supposed to work.

Well, yes. Wayland is a display protocol and it does its work splendidly. X11 does a lot more than display management, and those things should not be put into the display manager yet again as it is a mistake.

Maybe there should have been a unified api for things people have been wanting to do for 30 years instead of people figuring them out 12 years into development.

What degree of feature parity depends on the specific implementation. There are in fact a variety of things that don't in fact actually work or work only in a specific implementation adopted recently.

I'm presuming what you actually mean is that the specific features you use in insert environment on non Nvidia hardware work well enough.

I mean that everything you can do in X11 you can do in Wayland. That's feature completeness, right? The big issue seems to be that Wayland doesn't actually _do_ much. So everyone has to agree on how to handle copy/paste, screen sharing, etc. And sometimes Gnome and wlroots do things differently, and that causes problems. But it has nothing to do with the Wayland compositor not being feature complete.

If you are going to redefine feature parity X11 to mean feature parity with the much smaller core of features that you believe ought to be in wayland vs part of some external package like gnome/sway/kde it would seem that you can trivially argue yourself into deciding you have arrived no matter how impoverished the feature set you choose and how meaningless such a conclusion ultimately is.

Nobody wants to decide their interface based on a combination of gpu driver, apps, features are fully functional.

The actual question at hand is how many of the now more fragmented set of platform are feature complete and to what degree. Comparing apples to oranges is a waste of your time and my time.