Code review comment for lp:~vanvugt/mir/unocclude

> > I think you're planning for the wrong approach. The renderer should never
> feed
> > back directly to the client as it's just a /view/. The renderer shall never
> > communicate with clients in any way. Instead we should indirectly observe
> that
> > a client's become occluded. This is easy using BufferConsumingFunctor which
> > observes physical frame completion - if and when lp:~vanvugt/mir/judder
> lands.
>
> Hmm where did somebody indicate that the renderer should feed back to the
> clients?

"Compositors should not have the job of notifying clients of their decisions, the clients should get notifications based on things going on with their associated ms::BasicSurface."

Yes, it seems like the thing we agree upon is that nothing in the compositor system (GLRenderer, DefaultDisplayBufferCompositor, MultiThreadedCompositor) should do notifications of occlusion.

The big objection seems to be the future needs of other parts of the system noted by Alberto and Andreas and the loss of the logic by its removal.