[OSM-talk] Geofabrik Download Server Update

[Frederik Ramm]

Hi,

On 05/20/11 14:09, Dirk-Lüder Kreie wrote:
> You write there that the cutting is done sequentially from large to
> small. Seeing as how the polygon complexity is one major factor in the
> cutting process, wouldn't it make sense to use more smaller, more
> precise polygons to cut out the small stuff first with "better" polygons
> and then assembling the larger parts from that?

Not with the current Osmosis implementation. If you had, say, 150 
"precise" country polygons and were to cut them out of a 1.2 billion 
node planet file, then that would mean Osmosis does 180 billion 
point-in-precise-polygon tests, and that would take quite a while. The 
top-down approach I am using means fewer tests.

Better algorithms for tiling a surface exist but haven't yet been put to 
use in Osmosis. Very fast tile splitters exist in the mkgmap and routing 
preprocessing arenas but they only do rectangles and don't do 
--complete-ways which I am increasingly using.

Bye
Frederik