3 ms·

> Generally the worst offenders for Wayland support are closed source / abandoned apps that you should consider replacing. Why? Why does a working program hav

by bsder 3y ago

> Generally the worst offenders for Wayland support are closed source / abandoned apps that you should consider replacing. Why? Why does a working program have to change because the windowing system decided to flip everybody the bird? Windows doesn't do that. Linus himself will slam the wrath of God down upon anybody who breaks the kernel API. etc. I'll go further. The whole damn problem is the fact that there is actually a choice between Gnome, KDE, Mint, etc. Windows has one window manager. macOS has one window manager. Consequently, Linux has to wrestle with this stupidly horrible "Where is the dividing line between program, compositor, window manager, and video controller?" problem while other OSs simple put the problem where it belongs--in the damn OS. The issue is that Wayland forced a whole bunch of shit work onto app developers because the Wayland developers didn't want to expend the work to deal with it either. (Cue jwz's "Cascade of Attention-Deficit Teenagers" rant) Wayland is better. Take a look at copy-paste on X11 vs Wayland--well, actually, don't, X11 copy paste will make you throw up. Just use Wayland's copy-paste because you can actually figure out what your formats are. Wayland would be okay but while DeadRat is trying to force it on everybody they aren't putting in the work to make it work for everybody. And the app developers and users are quite rightfully salty and angry about all this.

anotherhue 3y ago
I'd agree with you if X was actually maintained, but we need to accept that it isn't. And to be fair, 'one app can read another app's data' is a reasonable thing to break. For everything else XWayland is the intended path (imperfect as it is).
- michaelmrose 3y ago
  It's actually incredibly useful to have apps like window managers or apps that rely on manipulating user input like xcape read another apps data. In fact such are still possible just 10x as complicated as they now involve writing awful little programs in some combination of c and yaml instead of a dead simple single line shell invocation of a program. See https://gitlab.com/interception/linux/tools https://gitlab.com/interception/linux/tools The fact that this can work at all seems to suggest that its just as possible for malicious code running as user to compromise your security and the only thing actually broken is useful apps.
- bsder 3y ago
  > I'd agree with you if X was actually maintained, but we need to accept that it isn't. Which is a self-fulfilling prophecy as the Wayland developers were the X11 developers. To be fair, I don't disagree with them. X11 was meant for a much different world. However, the Wayland developers have made some significant architectural mistakes. Fractional DPI for scaling is a clusterfsck. The memory ownership model is disastrous and makes communicating with languages other than C/C++ actively problematic. I can go on and on. Wayland, itself, needs more than a little bit of reboot at this point. Vulkan/DX12/Metal has won--Wayland should be built around those abstractions. WASM seems to be handling interop and bootstrap better and better--it should probably be based around that rather than DBus. Mouse/Touch/event/etc. management needs to be rethought and probably refactored as much as possible out of Wayland. etc. I suspect a lot of the problem is of the same origin, everybody wants to have unified abstractions for mobile/web and desktop. The Wayland/Gnome people do not want to have to write code twice--once for mobile/web and once for deksptop. In reality that turns into mobile/web abstractions for everything and desktop can go suck eggs.
  - charcircuit 3y ago
    WASM isn't an IPC mechanism. DBus makes much more sense to use.
charcircuit 3y ago
>Why does a working program have to change because the windowing system decided to flip everybody the bird? Because desktop Linux started off insecure and they are trying to tack on security to what exists. Keep in mind that xwayland exists so working apps will still work, they just can't escape the sandbox and interact with the rest of the system. >Windows doesn't do that. Windows has been trying to sandbox win32 apps for years. The difference is that Microsoft is made up of professionals who actually know how to develop and evolve an OS. >Linus himself will slam the wrath of God down upon anybody who breaks the kernel API. etc. No, Linus doesn't care about having a stable kernel API. Google had to develop a stable API for people to use. > Take a look at copy-paste on X11 vs Wayland Why are display servers handling copy paste in the first place? That is not where it belongs.
- michaelmrose 3y ago
  > Why are display servers handling copy paste in the first place? That is not where it belongs. It is fairly obvious that since your clipboard can hold very valuable info or be used to get you to do something that compromises your security that something privileged needs to intermediate between Application A wants to put something into a shared mailbox and Application B wants to take it out. Maximum security implies ensuring either action is done only by deliberate user interaction eg watching for real keystrokes or mouse usage. It also needs to be done in a standardized way that individual apps including apps you can't modify or update can work with so that an extremely basic OS function doesn't break. Can you explain why this doesn't belong in the display server and where indeed ti does belong?

by bsder 3y ago

anotherhue 3y ago

I'd agree with you if X was actually maintained, but we need to accept that it isn't. And to be fair, 'one app can read another app's data' is a reasonable thing to break. For everything else XWayland is the intended path (imperfect as it is).

michaelmrose 3y ago
It's actually incredibly useful to have apps like window managers or apps that rely on manipulating user input like xcape read another apps data. In fact such are still possible just 10x as complicated as they now involve writing awful little programs in some combination of c and yaml instead of a dead simple single line shell invocation of a program. See https://gitlab.com/interception/linux/tools https://gitlab.com/interception/linux/tools The fact that this can work at all seems to suggest that its just as possible for malicious code running as user to compromise your security and the only thing actually broken is useful apps.
bsder 3y ago
> I'd agree with you if X was actually maintained, but we need to accept that it isn't. Which is a self-fulfilling prophecy as the Wayland developers were the X11 developers. To be fair, I don't disagree with them. X11 was meant for a much different world. However, the Wayland developers have made some significant architectural mistakes. Fractional DPI for scaling is a clusterfsck. The memory ownership model is disastrous and makes communicating with languages other than C/C++ actively problematic. I can go on and on. Wayland, itself, needs more than a little bit of reboot at this point. Vulkan/DX12/Metal has won--Wayland should be built around those abstractions. WASM seems to be handling interop and bootstrap better and better--it should probably be based around that rather than DBus. Mouse/Touch/event/etc. management needs to be rethought and probably refactored as much as possible out of Wayland. etc. I suspect a lot of the problem is of the same origin, everybody wants to have unified abstractions for mobile/web and desktop. The Wayland/Gnome people do not want to have to write code twice--once for mobile/web and once for deksptop. In reality that turns into mobile/web abstractions for everything and desktop can go suck eggs.
- charcircuit 3y ago
  WASM isn't an IPC mechanism. DBus makes much more sense to use.

charcircuit 3y ago

>Why does a working program have to change because the windowing system decided to flip everybody the bird? Because desktop Linux started off insecure and they are trying to tack on security to what exists. Keep in mind that xwayland exists so working apps will still work, they just can't escape the sandbox and interact with the rest of the system. >Windows doesn't do that. Windows has been trying to sandbox win32 apps for years. The difference is that Microsoft is made up of professionals who actually know how to develop and evolve an OS. >Linus himself will slam the wrath of God down upon anybody who breaks the kernel API. etc. No, Linus doesn't care about having a stable kernel API. Google had to develop a stable API for people to use. > Take a look at copy-paste on X11 vs Wayland Why are display servers handling copy paste in the first place? That is not where it belongs.

michaelmrose 3y ago
> Why are display servers handling copy paste in the first place? That is not where it belongs. It is fairly obvious that since your clipboard can hold very valuable info or be used to get you to do something that compromises your security that something privileged needs to intermediate between Application A wants to put something into a shared mailbox and Application B wants to take it out. Maximum security implies ensuring either action is done only by deliberate user interaction eg watching for real keystrokes or mouse usage. It also needs to be done in a standardized way that individual apps including apps you can't modify or update can work with so that an extremely basic OS function doesn't break. Can you explain why this doesn't belong in the display server and where indeed ti does belong?