It’s still not ready, especially the KDE and GNOME situations. Things just aren’t very reliable. Mouse events are dropped and things like menus respond inconsistently. The XWayland layer doesn’t work smoothly, so X11 games and browsers aren’t very performant. Without that, there’s no transition layer, so it’s a cold-turkey jump to Wayland applications, of which there aren’t many.

I think starting fresh from a Weston or wlroots base is the only way we’re going to get decent desktop environments. GNOME and KDE are “translating” the X11 stuff over to Wayland methodology, but that’s not going very quickly.

It’s great that Debian is leaving the old X11 within arms reach. This is a very pragmatic way of pushing Wayland a bit while not screwing over the users.

I don't think you'll have much luck with Weston, it's not really designed to be used as a base (libweston is tacked on and the devs acknowledge that it's not great).

I've been using Wayland daily for over a year now, and have noticed none of these problems.

If I'm reading the announcement right, it's the default for GNOME alone, which makes sense:

GNOME now defaults to Wayland and the other desktops default to X, Debian has no reason to second-guess them. Worse, changing the defaults in this area would be a fork in all but name.

The problem with Wayland right now is that most (tiling) window managers need a very significant effort to be ported from X11.

Maybe sircmpwn could comment on this: Could wlroots (originally for Sway) make the transition process easier for other tiling window mangers wanting to port to Wayland?

Tremendously so, yes, but it's still going to be a lot of work.

Personally I’m waiting for a tiling Wayland WM that supports NVIDIA.

Effectively, you're waiting for NVIDIA to implement the same APIs every other Linux GPU driver uses.

You may be waiting a very long time.

Sway (Wayland equivalent of i3) works with the open source nouveau drivers, which seem to work quite well for some folks.

Wait. So I haven't used Wayland at all yet but why would the graphics driver need to have support for a tiling WM? isn't that in a whole different later and shouldn't matter at all?

Does NVIDIAs driver has any special code to make i3 work on X11 but missing code to make sway work on Wayland?

It’s the other way around: on X11, nVidia implements all the standard APIs, so window managers like i3 don’t need driver-specific code.

For Wayland, nVidia only implements the EGLStreams buffer management API, whereas other drivers implement the GBM API. Individual compositors (like sway) need to add support for EGLStreams.

See also https://wiki.archlinux.org/index.php/Wayland#Requirements

Note that Wayland doesn't have a full driver abstraction layer in userland like X does. Every Wayland compositior is expected to call into the Linux kernel drivers for the graphics card directly. GBM is just a thin wrapper that hides the fact that Linux never standardised the buffer creation interface across different hardware - it assumes that all those driver-specific interfaces wrap essentially the same shared Linux buffer management code, and it doesn't abstract stuff like modesetting at all. NVidia managed to reimplement enough of the Linux modesetting interface that existing code mostly works, but their own buffer management infrastructure is apparently different enough that they couldn't make it look like the one existing Wayland compositors are expecting to use. So they ended up creating their own slightly higher-level API instead.

Oh dear, so right from the start there is unnecessary complication hindering adoption of Wayland in yet another way. After reading your comment I assumed at least all the other drivers would implement EGLStreams too but following your link reveals no, so you couldn't even implement a simple compositor that only supports one.

Personally I avoid NVIDIA like the plague, but at work we have to support VMware running with 3d acceleration on all kinds of cards on Linux, including NVIDIA with proprietary drivers. Apart from VMware having announced no plans yet on supporting Wayland this seems like yet another small reason to stay with X11 until the end of time.

Xmonad[1] cannot be ported, but there is a Waymonad[2]. So yes, in most cases that would mean rewriting from scratch. What is worst of all this - no existing plugins/extensions/etc would work...

[1] https://github.com/xmonad/xmonad

[2] https://github.com/waymonad/waymonad

I would say "very significant" is understating it. Anything that touches Xlib/XCB/etc would need to be at least modified, if not rewritten completely.

How about gtk-2.0 apps?

That’s probably not a bad thing. We have the benefit of hindsight of decades of GUI management and it probably isn’t more than a weekend project to get a baseline library worked out to build cool things (like awesomewm and i3 and dwm) off of. Which probably already exists.

> it probably isn’t more than a weekend project to get a baseline library worked out to build cool things (like awesomewm and i3 and dwm) off of

what technical justifications are you using to make this claim?

Isn't that what wlroots is? And it's taken many years of work by multiple contributors and still isn't really finished

wlroots is finished enough to build production-quality Wayland desktops out of.

Will fluxbox be deprecated/compromised by Wayland? I've been happily using it since 2001 without a hiccup...

Yeah, aside from GNOME/KDE trying to do a smooth transition, it's kind of a breaking point for everything. We now have Sway as an i3 successor, for example. There are similar projects for other popular window managers, but from what I understand none of them are in a very usable state the way Sway is. This[0] page can serve as an informal list of all the Wayland compositors succeeding old window managers. Waybox is likely relevant to your interests.

[0] https://github.com/swaywm/wlroots/wiki/Projects-which-use-wl...

As I understand it, it was mostly a technical result of Gnome being the default desktop, and Gnome defaulting to Wayland.

I'm honestly astounded by how much fallout is generated by GNOME alone: GNOME strongly integrating with systemd pushing systemd itself[1] onto major distributions, now GNOME forcing Wayland.

Hell, RHEL even dropped KDE support entirely in version 8.

I'm not saying change is inherently bad, but I'm amazed how much change is caused by a single project.

[1] https://lists.debian.org/debian-devel/2013/10/msg00578.html

GNOME is not even that good even, not from a programming perspective (glib/gobject/gtk hell) nor from a usability perspective (features being removed left and right, high memory usage, etc). I personally think that tiling WMs do it way better, but even things like openbox and KDE are just more usable than GNOME. I am surprised that it is still the default in so many distros, something related to tradition i presume?

> something related to tradition i presume?

If only there were some sort of (M) Advanced Traditional Environment they could default to instead..

disclaimer: i probably haven't given any recent gnome version a fair shot

I doubt it; they already maintain patchsets for a lot of their packages, and one to change what gnome defaults to would be trivial. This was a deliberate move.

You can only see it as a “move” if you consider Debian to be a single reasoning entity. But consider: the person making such a patchset would be the Gnome maintainers, and they would not really have an incentive to do so; Gnome upstream probably changed the default for a reason, and the maintainers of the Debian Gnome package, being voluntary maintainers, probably are philosophically aligned with Gnome upstream. The rest of Debian, like the package maintainer who chooses what Desktop Environment should be the default, can not make such a change to the Gnome package without a quite adversarial process involving the Technical Committee, and in practice this process is rarely used, and even when used, rarely successful.

Therefore, it is the expected result that Debian should follow the upstream Gnome into using Wayland.

As a random sidenote, I find it very interesting that Wayland has pretty much all the same architectural decisions as Mac OS X 18 years later. Convergent evolution. It seems in general like the Linux world is a bit more stubborn to change.