[OSM-dev] Current Multipolygon definition of OSM considered harmfull!

[Stefan Keller]

I just stumbled over this page
http://wiki.openstreetmap.org/index.php/Relations/Multipolygon

I strongly recommend to reconsider this: Multipolygons are allright but...

1. Having zero to n holes (= inner boundaries) is an intrinsic property of a
'real' polygon according to computational geometry (c.f.
http://geometryalgorithms.com/books_top_10.shtml). Therefore that what's
described there should simply become a new basic geometry type called e.g.
'polygon' or 'area'.

2. Multipolygons in fact are a collection of 'real' polygons (where each can
have zero to n holes). It would be a huge source of confusion if you define
a Multipolygon the way it is currently.

3. It's better to define an own, clear XML encoding for such 'real'
polygons. Using relations to encode 'basic' geometry types like polygons
with holes makes the reading process of OSM files more and more clumsy.

4. If you really mean Multipolygon, I personally don't mind to use relations
for this (OGC's Simple Features in contrast did even define a basic type for
this).

Just my few cents... Stefan.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/dev/attachments/20080607/5df7e78c/attachment.html>