The server of static planetary data is actually just a simple http file server. Clients know what they need for a given viewpoint, and simply request the chunk if they don't have it already. The dataset is cut into pieces in such way that is (nearly) optimal for the progressive download of data that you need - a balance between the size of chunks and the overhead of bigger blocks.
The whole thing is just for convenience - we don't have to distribute the whole dataset, when in fact people will progressively use only a small portion of it.
