> And Wayland isn't going to do any audio output, though it'd be nice if there were a way to bundle together audio/graphic/dbus session addresses so that you could point an app at a coherent remote session by setting one environment variable instead of messing about with forwarding all of these protocols separately.
xpra.org deals with graphics and dbus notifications.
dbus was not designed to be networked however, so forwarding any more dbus based protocols would need to be done on a per case basis unfortunately.
As for audio, see my other reply: PA is not suitable without a fair amount of work (compression, support for disconnection, etc)
> Sane *async* protocol for cut-and-paste format negotiation/data transfer
xpra does that (mostly reasonably well too now)
> Custom cursor updating
xpra does that
> Communicating desktop geometry (including multiple monitors at different resolutions, etc.)
> Relationships between windows ("this window is a modal pop-up to that window")
xpra tries to do that (not easy - not even do-able in some cases since we cannot tell the client's window manager to "group" windows together so they move together - but we manage ok)
> Mapping windows to processes ("this process appears to have frozen; kill it?")
Not supported yet (hard I think)
> etc. There's just a ton of fiddly details like this.