Code review comment for lp:~mir-team/mir/touchspot-visualizer

>> I think it's potentially bad idea for minor stuff like touch spots to ever be part of the
>> scene.
>> You're committing yourself to the overhead of maintaining structures/objects for trivial
>> textures. That's certainly sub-optimal in complexity and performance.

So part of the in cursor and touchspot visualizations being a render-able is a performance based goal. A renderable representation makes it easy to allow hardware backends to render the input visualizations as overlays when appropriate (HWC, drm cursor API, or future DRM unified planes API), while still supporting a software fallback when required.

If we didn't have a retained representation of the input visualizations it seems like this could be really difficult.