Why isn't Xwayland just a Wayland client?

[Joseph Burt]

On Fri, Sep 8, 2017 at 11:02 AM, Pekka Paalanen <ppaalanen at gmail.com> wrote:
> On Thu, 7 Sep 2017 21:18:48 +0200
> Joseph Burt <caseorum at gmail.com> wrote:
>
>> Hi all,
>>
>> On Wed, Sep 6, 2017 at 1:09 PM, Daniel Stone <daniel at fooishbar.org> wrote:
>> >
>> > I really wouldn't recommend doing this.
>>
>> On Thu, Sep 7, 2017 at 10:05 AM, Pekka Paalanen <ppaalanen at gmail.com> wrote:
>> >
>> > I kind of wish I shared your optimism, but I'm thinking more of a death
>> > by a thousand papercuts kind of situation, not a single big failure
>> > point.
>> >
>> > I'd be screaming in delight and joy if someone proved it is possible to
>> > turn Xwayland into just a regular Wayland client that has no special
>> > privileges nor requires any nasty Wayland extensions.
>>
>> Alright, I'm hearing a consensus of: worthwhile if possible, but
>> probably unpleasant, somewhere between fiddly and very painful. I'll
>> hack a bit at the problem and see how much it hurts.
>
> Hi Joseph,
>
> exactly. Good luck, reserve a good stock of booze, and don't waste all
> of your sanity or liver! ;-)
>
>
> Thanks,
> pq

HI all,

This is absolutely possible. What I'm working to implement at the
moment: Every managed, user-movable X window is a surface, and is
alone and static in its own full-resolution "workspace," which makes
the surface-local to X coordinate transformation just adding a
multiple of the vertical resolution to y. This is to avoid having to
move fake positions around on resize. Position hints are ignored.
Everything else is a subsurface. I'm working to get this all into
Xwayland itself to avoid proxying. We'll see about that.

Can anyone think of a fatal gotcha here?

Cheers,
Joseph