Code review comment for lp:~robertcarr/mir/simplify-depth-assignment

> I'm still deeply suspicious that DepthId is the wrong abstraction - so this
> has the feel of "shuffling deckchairs on a sinking ship".

+1

I'd rather just present a z-ordering and have the shell have better tracking of what surface is where and how its stacked. I *somewhat* detect that the shell wants to group surfaces and order them more arbitrarily, and are trying to fit their square peg into SurfaceStack's round abstraction of a scenegraph

That being said, I suppose that if its already in the code base, its better that there's a way to use it.

My question is... are we planning on keeping DepthID forever, or do we think that as the shell evolves we will eliminate this? I sort of get the feeling that the stack would be a better abstraction if it was closer to a z-ordering (simpler).