Andrew,
Dang... we were hoping no-one would bring that up...
You're absolutely right, in a theoretical sense. If you need to process
global-scale data, the cleanest way to do it is to process the data
using pure geographic coordinates (this is usually called a "geodetic"
model).
The reason that a lot of systems still use a planar assumption is that
the algorithms for geodetic are substantially more complex to design and
implement, and there are even fewer sources for these algorithms than
for planar algorithms. Spatial databases are only just starting to
offer geodetic as a fully-featured datatype. Oracle, IBM and MS
SqlServer all provide this now (with varying degrees of usability).
The hope is that PostGIS will support geodetic as well at some point -
but unless some big sugar-daddy pops up with a whack of funding, this is
likely to be later rather than sooner.
J. Andrew Rogers wrote:
Isn't this largely a fundamental issue because, for good historical
reasons, we do geometry computations on 2-dimensional projections of a
complex object in 3-space? Your point is absolutely correct, but I
have always been interested in the reasons why geospatial is still
computed almost purely from projections in the computer age. (Not to
hijack the thread, but I have always been genuinely curious about
this; most of the underlying rationale seems suspect for an increasing
number of computer-based applications.)
Cheers,
J. Andrew Rogers
_______________________________________________
postgis-users mailing list
postgis-users@postgis.refractions.net
http://postgis.refractions.net/mailman/listinfo/postgis-users
--
Martin Davis
Senior Technical Architect
Refractions Research, Inc.
(250) 383-3022
_______________________________________________
postgis-users mailing list
postgis-users@postgis.refractions.net
http://postgis.refractions.net/mailman/listinfo/postgis-users
