Code review comment for lp:~vanvugt/mir/managed-surface

It's not a matter of having a superior solution. I'm trying to convince myself (or be convinced) that this is a solution at all for complex use cases involving metadata affecting multiple surfaces.

I don't disagree with shells wanting to attach per-surface metadata. I am more worried about metadata that is NOT per-surface. With this approach, such data would need to be replicated in each ManagedSurface instance. Also, there would have to be a "coordinator" to get the job done, which seems to be missing in this approach (or at least I'm not seeing it).

Perhaps you can provide a concrete example (code would be nice) of how such a complex policy involving multiple surfaces would be enforced. Perhaps a carousel WM that you can slide your finger to turn it to browse various app windows you have open. The surfaces would be rendered on the sides of the carousel. As the carousel is turning, I imagine you'd need to keep track of how much it has turned so you can continue to render your windows in their respective positions around the carousel's perimeter. But how much it has turned is not a per-surface info, and it's awkward to treat it as such.