4 ms·
> Wayland, itself, is just a protocol That is the root of the problems IMO. The thing is, this design decision had made the ecosystem a lot harder to grow than
by kbumsik 6y ago
> Wayland, itself, is just a protocol
That is the root of the problems IMO. The thing is, this design decision had made the ecosystem a lot harder to grow than X11/xorg.
- Zambyte 6y agoMaybe it has made is slower to grow, but harder? Now we have several dominant Wayland implementations including wlroots, Mutter, and KWin (the latter is still getting there from what I've heard). This is already 3x the number of dominant implementations of X11. Compositor developers can choose to use an existing implementation that suits their needs, or to develop one from scratch (that is compatible by implementing the same Wayland protocol). I think making Wayland just a protocol has made it easier to grow.
- floatboth 6y agoDominant compositor implementations – the dominant wayland implementation is libwayland :) About the only serious alternative is the Rust one – the crate has a native mode. (but you still have to use the libwayland backend if you want to do GL because of how Mesa's EGL impl is tied to… yeah, lol, hard problems here :D) But seriously, the "competition" between these compositors is awesome. This is like web standards again. Testing your client across different compositors reveals bugs (either in them, or in your client) and everything evolves together in standard ways. With the single Xorg server, everything got completely ossified, if anyone wanted to rewrite a full "production ready" X11 server they'd have to be bug-compatible with Xorg.