wayland vs xorg

Lacike · 2024-01-13 23:07:59

I have decided for arch in order to improve my linux skills and after upgrading my PC I have a brand new arch installed (new SDD). On my previous installation of arch I had Xorg with Awesome. Now I am considering to give a kick to a Wayland, as there are some vulnerabilities related to Xorg reported, but does not seem to be a critical issue. Is it worth to switch to a Wayland instead of the Xorg? And also, what is the best solution for a tilling window manager with Wayland?

Thanks in advance.

Trilby · 2024-01-13 23:33:51

No one can possibly answer this for you. Try both and see which you prefer.

The wiki has a list of tiling compositors. There's no such thing as a window manager in wayland.

Note that I've been fully on wayland for quite a while and I've written my own compositor. Most of the supposed "advantages" over Xorg are bullshit. The biggest argument for wayland had always been that Xorg needs lots of "extensions" for tasks like transparency and / or compositing while wayland doesn't. This is true only to the extent that wayland doesn't need extensions for those tasks but it does for nearly every other thing that is taken for granted in xorg. The number of non-standard "protocol extensions" used by any given compositor is astounding: just for simple things like a taskbar even.

On top of that, for a while I had been assuming that I screwed up making my compositor as it had a weird recurring issue of freezing for ~10s at a time when certain programs were opened (or when they opened new windows). It turns out my compositor handles this condition better than many others. Other wayland compositor I've tried just crash completely in those moments. Specifically, if a wayland client seg-faults, many compositors will segfault and go down with it (particularly if the client is a layer-shell or fullscreen client).

This is actually a result of assumptions baked into the wayland protocol(s): for example in conditions in which the protocol requires that a client "must" respond in a certain way, it seems many compositors take that response for granted. If the client seg-faults (or is just not properly implementing the protocol on it's own end) many compositors do not check before assuming the client has done it's part. And moving on without checking results in seg-faults in the compositor code. This is astoundingly stupid, and I have never heard of something even half as fragile under X11.

(edit: arguably this could been seen as a fault not in the whole protocol, but rather in half-baked implementations. But in reality the protocol itself creates these "tripping hazards"; while it's true that a impeccably coded compositor might be able to avoid tripping in such situations, the protocol is still contributing to these very fragile conditions. This is perhaps most true of the layer-shell protocol which is needed for many things users take for granted, but the protocol is really just horribly written.)

X11 has a lot of cruft due to being written a long time ago for much older hardware. But it was written by very smart people who designed it well within the constraints they had. Wayland thows off all those constraints ... but also all the good design.

Last edited by Trilby (2024-01-13 23:48:30)

Lacike · 2024-01-13 23:37:50

Great, thanks for the info. I think I will stay with Xorg as for now. Works great and I have still plenty of stuff to learn with Lua. I will check the Xorg vulnerabilities, should be just fine.

Arch Linux

#1 2024-01-13 23:07:59

wayland vs xorg

#2 2024-01-13 23:33:51

Re: wayland vs xorg

#3 2024-01-13 23:37:50

Re: wayland vs xorg

Board footer