Label placement on the server

[kkaefer]

Since we already load the map stylesheet before delivering the tile, we can just as well do the label placement on the server in case this turns out to be expensive.

[ansis]

Currently, placement is slow in dense areas, and rotation collisions made it worse. But I think there is plenty that can still be optimized, and I'm optimistic about getting performance to the point where its reasonable. The perceived slowness could be largely solved by not waiting for text placement to render other features.

Doing label placement on the server would mean redownloading data when font sizes change. Also, in the further future, something like iD would need placement in the client so it can handle changing data.

@kkaefer do you think its worth working on improving this? Or do you think placement on the server is inevitable and we should just skip straight to that?

[kkaefer]

Nope, I see label placement on the server as a last resort. We should try hard to do this on the client.

[ansis]

Closing. Client placement needs more work seems doable.