On Tue, Oct 28, 2008 at 8:24 PM, Freek <[EMAIL PROTECTED]> wrote: > I have not seen those (would be interested), but I would guess rectangular > queries make an equilateral-triangle subdivision inherently less favourable, > even though the geometry is distorted by the projection
i think it might have been this one... or maybe not... it was a long time ago :-) http://research.microsoft.com/research/pubs/view.aspx?tr_id=736 > (we don't have much > data near the poles anyway ;-) this is exactly why an icosahedral decomposition is so good - each tile in the same level of the tree is very nearly the same projected area. the quadtile approach (whether using 3395 or 4326) projects to much smaller areas near the poles than the equator, so wastes valuable coordinate space and unbalances the tree. if the icosahedron is oriented correctly (fuller's dymaxion orientation) then several of the triangular faces are completely filled with ocean and can be omitted. cheers, matt _______________________________________________ talk mailing list [email protected] http://lists.openstreetmap.org/listinfo/talk
