[OSM-dev] osm2pgsql and only-named multipolygons

[Ben Supnik]

Hi Y'all,

I hope that Richard is right and the glorious future of area primitives 
is coming...Viva La Revolución! :-)  In the meantime...

It seems to me that multipolygons represent a chicken & egg 
problem...because the data quality issues are semantic (that is, you can 
create legal OSM XML that is nonsensical when interpreted)...

- If there is weird tagging in the DB, the tools have to support weird 
tagging.
- If the tools support weird tagging, users will continue to create 
weird tags.

(By weird tags I mean any tagging scheme for a multipolygon that doesn't 
put the data for the multipolygon _on_ the multipolygon. :-)

So I support Richard's notion of making Potlatch "more picky", even 
though my own code (which strictly interprets the data, but does not 
create new data) has to deal with all of the stuff that the map already 
renders.  If a user is going to take the time to change the data, force 
them to put the tags on the multipolygon itself..it'll make life better 
for everyone in the long term.  Even the unambiguous uses of 
tag-on-outer-way are ambiguous.

The one thing I wonder about: _if_ a large number of clients (including 
the major maps) could all agree on a _limited_ definition for 
"promoting" outer-way tags to a multipolygon, one that was less 
forgiving, perhaps this would also encourage the community to not tag 
data in ambiguous ways.

The tools I use to build X-Plane scenery currently promote _any_ 
outer-way tags to a multipolygon that doesn't have its own tags...this 
is way too permissive, based on the comments others have posted to the list.

cheers
ben