There are a number of usecase that require extracting individual OSM data. For example, we may want to draw an outline of a city, and the browser should be able to get this data from Kartotherian. My understanding is that this data is stored in the OSM as a relation. I was told that OSM IDs are not very stable, so I think wikidata IDs would provide a much better IDing mechanism. Also, I think we should target topojson format from the start, as it will offer a significant bandwidth optimization.

Code is at https://github.com/kartotherian/kartotherian-geoshapes