[postgis-users] Points or boundaries more efficient for querying?

[Nick Bower]

Hello.

We have large quantities of satellite metadata that we will reduce to a
set of standard tiles covering a large global region of interest.  We
are doing this because storing the data at native resolution will mean
an enormous amount of irregularly located data to query later.

We are to spatially query the data; eg give me the tile corresponding to
my location, and then, give me records for this location corresponding
to my metadata criteria (ok normally this would be a join on a single
query I guess).

So what, is the most efficent way to represent the geometry information
(the tiles); as centre points or tile boundaries?  I initially thought
points and queries using minimum distance would be least redundant, but
then the rtree documentation seemed to point to using regions,
boundaries and contains/intersects type queries would lead to better
index performance.  Is this the case?  The tiles will of course be
regular and adjacent.

Ta, Nick
-- 
  Nick Bower
  nick at nickbower.com