On Tuesday 28 October 2008, Nic Roets wrote: > Well, I'm really pleased with the performance of the fixed tile size > algorithms I implemented in gosmore. OSM data is remarkably "flat" : Either > a tile is empty (sea / rural / unmapped) or has approximately the same > number of objects as any other urban tile. So if the tiles are small enough > you don't need a second index (e.g. subdividing the tiles).
Interesting observation... > * Arrange the tiles in a 2-D Hilbert curve to reduce disk seeks Still, I think the idea of Asano et al. [1] applies because you are retrieving a set of square tiles within a rectangular query region (with a reasonable aspect ratio), but I don't know if it's really noticeably better in practice. [1] Publicly available version: ftp://ftp.inf.ethz.ch/doc/papers/ti/grpw/SFC.ps.gz -- Freek _______________________________________________ talk mailing list [email protected] http://lists.openstreetmap.org/listinfo/talk
