Implementation possibilities

Cases:

• Rendering map tiles for use in Leaflet. i.e. given x, y, zoom
• Rendering static map images of arbitrary size for embedding/external uses; e.g. Walking Papers
• Rasterizer for UTFGrid generator

Best scenario we make it modular so that it asks for all the geodata within a particular bounding-box and it doesn't care whether that comes from Postgis, mongo, or even just all the points from a shapefile. In other words, it shouldn't fail if we give objects outside the bounding box but we might not want to spend much time optimizing the data that is passed in.

It just needs a hook to be able to ask your application code what geometries you want to render. The naive implementation (our demo) just returns everything in that hook. In a real situation, you'd have it all in a spatial DB anyway; this hook is where you'd do your query.