Xfce 4.18 Planning Begins With An Eye On Wayland Application Support

[oiaohm]

Something like wlroots is really necessary, yes! What is it for a bad way to keep the number of DEs and WMs small or focusing everthing on Gnome by complicating the
development process or bring it back to the 80s? Even big projects need years to support Wayland. One solution, take it or die is everything against what OSS stands for.

Take it or die has been the x.org X11 server or the Xfree86 X11 server before it. Video card vendors have not been putting up drivers to support multi development of display servers.

The past prove a bad way too. Noone have a idea in which way Gnome will change Mutter in future. Maybe they find out windows are too complicated for users? They proved they are ready to do radical steps and ignore input from users or other projects. I would they forks of Gnome have much more users as Gnome themself, thats say all.

This here ignore that a lot of open source items exist to suit the developer/funding parties needs. Redhat targets enterprise based customers in that market a univerial do everything desktop does not have to be there. Think using a mouse with a car entertainment system.

[You-]

I dont think it would be useful to team up with Biudgie - that project (and the distro, I forget its name) is also more or less dead.

It will be interesting to see if they choose the wlroots or the mutter path. But even without the desktop moving to wayland yet, the apps having support is a huge step.

[curfew]

The past prove a bad way too. Noone have a idea in which way Gnome will change Mutter in future. Maybe they find out windows are too complicated for users? They proved they are ready to do radical steps and ignore input from users or other projects. I would they forks of Gnome have much more users as Gnome themself, thats say all.

I think Mutter is pretty generic a window manager, the only issue is this single dependency to a Gnome-specific package. If I recall right, I did actually personally trace the dependencies and despite the name "gnome-desktop" it is more or less a meta package which doesn't pull anything substantial from the actual Gnome desktop. (Some pointless weather widget for the lockscreen or something like that.)

Mutter isn't same as Gnome Shell, where most of Gnome's look-and-feel originates from.

I am yet to try the lambasted Gnome 40 but I can see it lurking in the Arch repos already. XFCE might win a few Gnome-escapees on their side if they just were a little bit quicker with this transition to Wayland... I am hoping they will choose Mutter because it could make using modern Gnome apps on XFCE little bit nicer.

I have also been saying for a long time that I would like to see Mutter become that standard implementation for GTK-based desktops looking to support Wayland. I don't like fragmentation and do not believe that libraries alone are sufficient to avoiding it. No, we need a concrete display server-like master process that all DEs will use. That could be Mutter.

[DanL]

CSD, Wayland, Mutter. Excellent choices.

Someone please ban this troll.

[oiaohm]

I have also been saying for a long time that I would like to see Mutter become that standard implementation for GTK-based desktops looking to support Wayland. I don't like fragmentation and do not believe that libraries alone are sufficient to avoiding it. No, we need a concrete display server-like master process that all DEs will use. That could be Mutter.

This is not as simple. With the New Nvidia binary drivers coming with DMA BUF will equal less work for those who want to make a Wayland compositor. Why DMA BUF buffer pass around will be able to be used as the default way no secondary code paths just to support Nvidia.

There is a fragmentation problem with the way Nvidia provides their binary drivers compared to open source drivers that really does need to be addressed.

Remember the problem of IPC overhead and scheduler mistakes. Libraries method of Wayland avoids having to use IPC between a display server and windows manager stuff.

[angrypie]

<birdie's usual tiresome drivel>

If, for every anti-Wayland propaganda message you post here, you contributed two lines of code to make Xfce Wayland-ready, they'd be better off than even GNOME.

[curfew]

Remember the problem of IPC overhead and scheduler mistakes. Libraries method of Wayland avoids having to use IPC between a display server and windows manager stuff.

I don't really mean that there should be a separate display server application that window managers talk to via some IPC protocol. Just that there is a single application instance that isn't patched for every single desktop environment separarely. The master instance itself could even be a library but functional in itself without any complex bootstrapping or initialization in client-side code.

[Vistaus]

They either want to use Mutter or upgrade their existing compositor with wlroots.

You sure? 'Cause I have read in the Phoronix comments time and time again that they will use Mir to move to Wayland.

[Danielsan]

I vote for wlroot!

[oiaohm]

I don't really mean that there should be a separate display server application that window managers talk to via some IPC protocol. Just that there is a single application instance that isn't patched for every single desktop environment separarely. The master instance itself could even be a library but functional in itself without any complex bootstrapping or initialization in client-side code.

That is what wlroots and libmutter are both working in the direction of. The expanding functionality in libwayland-server from the reference wayland compositor is also heading in that direction.

Something to remember here the fragmentation of buffer handling between Nvidia closed source and Open source drivers has resulted in way more complex bootstrapping/initalization and needing more than 1 code path as graphical operations that will work on all the open source drivers will not work on Nvidia eglstreams offering of the past.

Big thing to make writing wayland compositor simpler so the reference libwayland-server can do more with less bootstrapping code is get Nvidia closed source to toe the line and provide the same interfaces everyone does. Nvidia with direct X under windows has to provide interfaces that suite what Microsoft wants we need the same with wayland. The days of releasing custom driver for the display server so you can be a unique beast to everyone else is coming to the end.