News for hackers. Stuff that matters.
Re:“We currently have no plans to support Xwayland” by hyperion2010 on Thursday July 20, 2017 @11:31PM (#14817850)
I've been following the X11 vs Wayland story since very close to the beginning and have built and used wayland and weston multiple times over the years. Cutting through all the FUD on both sides, my conclusion is that Wayland was created in the context of some very specific use cases which essentially preclude it from being a general solution that can replace X by itself (it could succeed if there was a single project to reimplement X in a sane way, but leaving it to GKT and QT is NOT a solution by any stretch of the imagination). Nvidia's engagement with linux simply does not cover the use cases that Wayland was created to address. In my view desktop/workstation use cases are not sufficiently covered by Wayland, and that is where nvidia's focus has been.

tl;dr Nvidia has no need to play in this space, anyone using X will just use X without the translation layer.

Re:“We currently have no plans to support Xwayland” by madez on Thursday July 20, 2017 @11:40PM (#14817892)
While your post is interesting, it is also vague. Can you specify what Wayland is missing that can't be adequately handled by QT or GTK, or do you just have the feeling there is such a thing?
Re:“We currently have no plans to support Xwayland” by hyperion2010 on Friday July 21, 2017 @12:49AM (#14818135)
Without speaking too much FUD from the X11 side the main issues facing workstation use cases are input handling and permissions. The current wayland party line is "just use dbus to communicate between windows." This is complete madness because it basically means that if you want programs to be able to work together and not steal each others keybinds they all have to depend on a message passing framework. Full disclosure I have done everything in my power to avoid installing dbus and having dbus dependencies, but it is getting harder and harder. That said, it doesn't matter what you think about dbus, it is completely unrealistic to expect that every piece of gui software written for linux to be rewritten to include dbus. This is basically going back to the days of cooperative multitasking and I think we all know how that ended, the overhead is just too high.
Re:“We currently have no plans to support Xwayland” by wuch on Friday July 21, 2017 @03:44AM (#14818768)
What exactly do you mean by "stealing each other keybindings"? From what you have described so far, it would seem to me that you are describing problem specific to X11 where global keybindings are often implemented in spyware manner, i.e., listening to all keyboard events (regardless of current window focus) and reacting accordingly. They may conflict with each other, they may interfere with each other, they may spy on user, in other words they are not working together at all.

There is no standard way to register global keybindings under Wayland yet, but this is bound to happen eventually. Not as a builtin part of Wayland, but it doesn't matter. It will of course use D-Bus interface, as KDE does it now for example - but if an application doesn't support this interface, it can't steal keybindings, it just doesn't get to install global keybindings itself.

Not sure what you have exactly against D-Bus, but clearly modern Linux desktop embraced it.

Re:“We currently have no plans to support Xwayland” by gens on Friday July 21, 2017 @10:02AM (#14820593)
I have something against dbus. Bout the protocol and the implementation are very much over-engineered. The protocol at least has an excuse.

In the desktop things, that wayland aims to do, dbus is useless. (actually i don't know what it's useful for at all, other then quickly hacking some object oriented (gui)client-server desktop program)

I do recommend reading the dbus protocol (the protocol, not the API).

Sorry to sound negative but "modern linux" now means "desktop crap", and it is all crap. Freedesktop.org (the linux "desktop" "standards" authority) has gone from "ehh, passable" to "complete crap" in the last... idk 5-10 years (the "desktop" quoting is because it has gone pass the actual desktop problems).

Re:“We currently have no plans to support Xwayland” by sjellis on Friday July 21, 2017 @03:21AM (#14818684)
I only follow Linux graphics casually, but AIUI one of the principles of the Wayland initiative was to replace the monolithic model of X with a combination of more tightly scoped components. For example, libinput was created to take responsibility for actually handling input devices, so that Wayland compositors would never need to carry that code, and the input device handling in X could be decommissioned.

From that point of view, "Wayland" is kind of a short-hand for a number of different stacks that have some components in common, and speak the Wayland protocol. If you choose not to use a complete stack on a system, then the result can't be fully functional.

Personally, I don't have any sense of how much dark matter is in the area of Linux GUI applications. We can easily know about Open Source, where most modern applications sit on toolkits and desktops that are moving to Wayland, but presumably places like CERN have some really old internal applications that will only ever run on X until they are replaced. I'm genuinely unsure how large that problem is.

I have to say that I'm always skeptical about arguments that something can't be done for performance reasons. People have been using that argument against new technology since at least the days when C was decried as new-fangled, decadent luxury.

Re:“We currently have no plans to support Xwayland” by vertex-four on Friday July 21, 2017 @08:17AM (#14819759)
Can you provide an actual example of what is impossible without dbus? Getting mouse/keyboard input for your window is part of libwayland, and not dependent on dbus at all.
Re:“We currently have no plans to support Xwayland” by microcolonel on Thursday July 20, 2017 @11:44PM (#14817914)
> (it could succeed if there was a single project to reimplement X in a sane way, but leaving it to GKT and QT is NOT a solution by any stretch of the imagination).

Qt and GTK are not responsible for implementing the compositor at all. Compositors aren't actually the problem at all, the problem is applications.

XWayland is one of the main ways we are going to bring legacy applications to Wayland desktops. It works about as well as X does, and lets the other applications take advantage of the sane architecture of Wayland.

> In my view desktop/workstation use cases are not sufficiently covered by Wayland, and that is where nvidia's focus has been.

What use cases? The only thing that's missing for me is input method integration. It has full display colour management, handles multiple refresh rates in the same session well, handles scaling, supports every input device I own (including graphics tablets). What's not really firm right now is native support in every application, that's coming along at some rate, and XWayland is bridging that gap quite well.

Re:“We currently have no plans to support Xwayland” by hyperion2010 on Friday July 21, 2017 @01:12AM (#14818215)
Qt and GTK and responsible for implementing how all of their windows talk to each other. My point is that it is not reasonable to expect or require programs to depend on a widget toolkit just so they can get proper keyboard input. I realize this is a bit of an exaggeration, but X11 is more than something that draws windows and from my understanding wayland doesn't address many of the other parts. For example, from my understanding wayland does not provide a standard way to support window manager level keybinds that override the binds set by the developer of the window that currently has focus. For me this means that the primary way I interact with my computer is now broken on wayland unless every single program I use implements the ability to pass those custom keybinds to the program that actually handles them. Sure, everyone can implement their own way of dealing with window manager level keybinds but that completely defeats the purpose of having a standard in the first place.

Probably the most salient issue with wayland is that it prevents programs from communicating with each other in the name of 'security' and 'sandboxing' when preventing those things at the display server level adds literally zero security and just makes everything more painful for application developers (imo to the point where they simply won't touch the stuff). If there were a 'windowing protocol' that went along with wayland that covered many of the things wayland does not then I do not think there would be as much of an issue, but the thought that we can force a new display server on people without the other bits as well seems at best naive. Again, I think that the absence of such a 'windowing protocol' from wayland is because one of the main use cases it was developed for often had literally one program running (car backup camera displays).

I do not dispute that xwayland bridges the gap. But if people are using X right now then nvidia has no reason to support xwayland right now because there are no wayland only applications that their customers are demanding support for. Down the line that may change, but from a corporate decision making point of view there is zero reason to do this.

Re:“We currently have no plans to support Xwayland” by microcolonel on Friday July 21, 2017 @03:49AM (#14818793)
> Qt and GTK and responsible for implementing how all of their windows talk to each other. My point is that it is not reasonable to expect or require programs to depend on a widget toolkit just so they can get proper keyboard input.

You don't need to. Just like X11 has Xlib and libxcb, Wayland has libwayland. It is actually easier to get input from libwayland than it is to get it from Xlib, and about on par with libxcb.

> I realize this is a bit of an exaggeration, but X11 is more than something that draws windows and from my understanding wayland doesn't address many of the other parts. For example, from my understanding wayland does not provide a standard way to support window manager level keybinds that override the binds set by the developer of the window that currently has focus.

You're right about that, global keybindings will have to be allowed through the compositor, and I don't know if there's a protocol for that yet. Global keylogging is one of the things Wayland was designed to avoid, so the way we implement that feature will have to be a bit different. It's for good reasons though.

> For me this means that the primary way I interact with my computer is now broken on wayland unless every single program I use implements the ability to pass those custom keybinds to the program that actually handles them. Sure, everyone can implement their own way of dealing with window manager level keybinds but that completely defeats the purpose of having a standard in the first place.

I don't know that it defeats the purpose of a standard! I agree that these use cases need to be addressed, but I don't think the way to do that is to give every Wayland client the ability to globally change the way input reaches other clients without permission from the compositor.

> Probably the most salient issue with wayland is that it prevents programs from communicating with each other in the name of 'security' and 'sandboxing' when preventing those things at the display server level adds literally zero security and just makes everything more painful for application developers (imo to the point where they simply won't touch the stuff).

Yeah, I think that the XDG compositor extensions should loosen this stuff up a bit, or at least make it easy for an application to ask the user for permission to spoof everything. I'm unlikely to be displaying a hostile application on my machine. That said, doesn't exactly add "zero security" though. It is helpful for partially-isolated applications, like those in containers. I would like to be able to trust the isolation of applications some day so that I can run spotify on my workstation without breaking into a sweat.

I think it would be perfectly adequate to have the application ask for permission to do the heinous boundary violations that make X11 fun and dynamic.

> I do not dispute that xwayland bridges the gap. But if people are using X right now then nvidia has no reason to support xwayland right now because there are no wayland only applications that their customers are demanding support for. Down the line that may change, but from a corporate decision making point of view there is zero reason to do this.

For what it's worth, NVIDIA is looking to support Wayland on desktop. They already support it commercially in their embedded products (especially targeted toward automotive). So it's not exactly accurate to say that they have no customers asking for Wayland-only application support. The question is, when will it be convenient on desktop Linux? and that's hard to say. It is somewhat surprising that they aren't currently planning to support acceleration in XWayland, given that they have at least put some effort into supporting Wayland in general.

Re:“We currently have no plans to support Xwayland” by ryanmarsh on Friday July 21, 2017 @12:10AM (#14818004)
tl;dr it's 2017 and video is still a wasteland on Linux