Re: Wayland plans for gnome-shell? (was Re: Unity Update)

[Owen Taylor <otaylor redhat com>]

On Fri, 2010-11-19 at 12:07 +0100, Piñeiro wrote:
> From: Alan Coopersmith <alan coopersmith oracle com>
> 
> > Piñeiro wrote:
> <skip>
> >> I know that probably this comment is doesn't 100% suits in this Unity
> >> thread, but as you are talking about the GNOME 2.0 timeframe, I think
> >> that it is worth to mention that for GNOME 3.0 time frame, it is
> >> supposed that the magnification will be a feature instead of a
> >> separate a11y tool. But a feature of GNOME-shell [1]
> > 
> > Good, sounds like what I was saying about having a11y features in the normal
> > compositor instead of a a11y-only one like gnome-magnifier.   But for it,
> > that still leaves the questions of:
> >   1) Is gnome-shell ever going to port to Wayland or only support X11?
> >   2) Does Wayland provide everything gnome-shell needs to provide those
> >      features?
> > 
> > Wayland changes the way compositors integrate with the window system,
> > so work would probably be needed to make gnome-shell work well with it.
> 
> No idea of Wayland plans for gnome-shell. CCing gnome-shell list so
> they can provide extra information if they want.

I think we'll eventually go to Wayland - that's the way that things are
moving in the X world. But it's not going to happen tomorrow.

Reworking gnome-shell to output to the hardware directly via EGL and act
as the Wayland compositor is not a huge job. I bet someone motivated
could get a demo up in a week.

However, there are a some issues that going to take longer:

We've spent a long time starting from what was basically a bare
windowing system and building a desktop on top of that - a lot of things
need to be replicated in Wayland - application <=> window manager
interaction, drag and drop, cut and paste. Etc. And they don't have to
just be replicated, they need to be made to interoperate reasonably well
between native Wayland apps and X apps.

And hardware driver compatibility issues. I think we should be able to
pick a direction for Linux graphics and expect NVIDIA and AMD to catch
up with it for their binary drivers, but when that's going to require a
major re-architecting of the way that their drivers work on Linux, we do
have to give some advance notice and give them things to test with
before we flip the switch.

There may be some significant performance work to be done on the open
source drivers as well... as a random example, the Intel 3D driver is
really aggressive about caching allocated memory. That's fine for a
single 3D game or even a single 3D compositor, but if every application
started doing that, there could be a pretty big jump in memory usage.

GNOME 3.4 in 2012 is the earliest I could really imagine us switching
over and even that might be a bit ambitious (we have a lot of other
stuff to do for GNOME 3.x too!)

- Owen