Client display configuration notification

[raof at ubuntu.com]

On Tue, Oct 20, 2015 at 11:16 PM, Gerry Boland 
<gerry.boland at canonical.com> wrote:
> Hey folks,
> it appears this topic has gone stale, without any consensus being 
> reached. A core issue impacting this discussion which remains 
> undecided is the following:
> 
> - Is it in Mir's scope as a display server to provide a client api 
> for configuring mir server settings like those for displays & input 
> devices? Or should Mir expose a mirserver API for these settings, and 
> require the shell to implement its own API via an alternative channel 
> (e.g. DBus) for a settings client to use?

I have a well-known aversion to shipping API that is useful for exactly 
one client which is always shipped with the shell and is harmful if 
used by any other client. Both the set-base-display-config and 
set-input-configuration APIs fall under this concern.

Now, that said, this *is* something that every shell is going to need 
to handle in some fashion, and it's perfectly reasonable for Mir to 
provide support for it. In the Wayland world this is just done with 
ad-hoc protocol, because that's easy. For us, I think that having a 
separate libmirclient-shell¹ that contained these APIs would satisfy 
my paranoia and also be useful for Unity8.

For bonus fun points this will be another capability check needed on 
the MirConnection. Can we add a create_privileged_session_for(socket) 
API, or is it still not possible for Unity8 to hand sockets off to 
clients?

¹: Name to be exhaustively bikeshedded.