Understanding the working of X11 from the ground up (2017)

Understanding the working of X11 from the ground up (2017)
magcius.github.io

I'd say the biggest issue is really maturity. I actually find Wayland to be a surprisingly nice piece of software engineering. I don't know if I like the architecture really, but it comes together fairly nicely in the end.

For example, there's unanswered security questions. It would be good if applications couldn't just go and manipulate and spy on other apps like they can with X11. Wayland indeed solves this issue, but it leaves other problems. How do I capture the desktop?

wl_roots offers some extensions that help alleviate the problem, but they don't solve the full problem; for example, you can capture the screen, for which programs and an OBS plugin exists, but there's no upstream Wayland extension that can make WebRTC desktop capture work. I think what we really need is the ability to integrate policy and consent requests, but I'm not sure whether there's progress on that front, or how long it will take. Once the problem is solved, upstreamed to WebRTC, and that upstream version trickles down to Electron apps, Firefox and Chrome stable versions... well, things will be in a better place.

Speaking of web browsers, today Firefox still has some issues to go before it is fully working in Wayland natively. You can do it today, but it is not perfect; there are sometimes places in the code where it assumes it is running on X11, and things simply break. The screenshot tool is just flat out not present when running under Wayland due to the above issue.

Applications are slow on the uptake too. XWayland covers most of them, but it is not perfect. Like anything that captures the screen, of course, is not going to work under XWayland, but also, XWayland apps may behave a little off. In Sway WM, one bug is that when the mouse travels from an X window to a Wayland window, sometimes it behaves as though the mouse is still hovering over the X window at the last position it touched, which leads to tooltips showing up where they shouldn't. There's still a decent amount of GTK2 apps, and even Qt apps aren't all Wayland yet; Krita plans to start working on support I believe next year.

In my case, here's the itch I'm currently scratching: Sway WM also does not have support for drawing tablets. I found this out, by, well, loading up a drawing program and noticing the lack of pressure sensitivity. So, that's what I'm currently working on. Needless to say, this is a non-starter for anyone that relies on a graphics tablet, although if you are just using GNOME 3 it is unlikely to be an issue since I'm sure GNOME 3 supports graphics tablets just fine.

NVIDIA drivers are still a pain in the neck. I gave up and bought an AMD graphics card, and frankly I can't recommend it enough. Basically everything works better on Linux, but certainly Wayland works better. It's worth giving up some raw power. Still, for many people the proprietary NVIDIA drivers are going to need to work, and today support varies from lackluster to poor depending on what compositor you use.

Wayland is a protocol for a well designed future where all that functionality is broken up into separate pieces of software. Wayland handles the window model and most of the other parts already live in their own software projects. The problem is that Wayland makes some changes to the basic window model on the way through that are a big deal and so far there is no great software written that uses it.

Adoption is very slow because the best case scenario is there a Wayland compositor that has feature parity with an X11 window manage and a bunch of weird problems cropping up because screen recording is a to-be-solved problem.

My position from the armchair is that Wayland really needed a complimentary official protocol for how to request another application's graphical output. Not coordinating that means that anything that supports Wayland is going to be practically inferior to an X11 solution. It'll get fixed sooner or later I expect, it isn't a technically hard problem.

There will be people who don't want to use GNOME frameworks and don't want to use the GNOME window manager.

Wayland is enforcing a policy ('don't access other devices graphics output') where it should be providing capability ('an applications graphics output is private by default but that can be configured'). However it looks like the design of Wayland purposefully omitted that sort of configuration from the protocol. So instead of having a security model that is sufficiently complicated to allow screen recording, the security model is 'No'.

To get a framework independent solution all the compositors would need to implement an identical API for accessing graphics output. That needs to be coordinated in the same fashion as the Wayland spec. As the spec stands, it is just asking for every framework, compositor or application to go implement its own different solution.

The Wayland designers may even have tried to avoid explicitly creating a security model, but in practice it comes with an implicit security model that is inflexible, unconfigurable and difficult to coordinate around.