[OSM-dev] Tag Handling in the Database

Dave Stubbs osm.list at randomjunk.co.uk
Thu Mar 27 03:58:26 GMT 2008

2008/3/27 Mikel Maron <mikel_maron at yahoo.com>:
> The discussion that followed Gervase's email last week was fascinating,
> especially the analysis of the OSM community as a anarcho-collective.
> However I was surprised to see no one suggest necessary (at least to my
> mind) technical solutions that don't even touch on the subject of the tag
> "approval" system.
> Currently there's massive duplication of information on tags. Besides what's
> described below in the wiki, there are presets defined in JOSM and Potlatch,
> rendering rules for osmarender and mapnik, and other places as well.
> Can the key/value, rendering rules and outputs, comments and descriptions,
> photos, and perhaps if we decide it makes sense, the "approval" process be
> abstracted to a model for use in rails? What would the goal look like, and
> are there small intermediate steps which are obtainable?

I kind of did suggest it. But only in a general sense without anything
actually useful like some code.

The main problem we have is the disconnect between what is (the
currently used tags), and what we want. I'd favour some technical
solution involving analysis of the current planet file and some
integration with wiki-docs and a trust-based scoring system.

This isn't exactly an easy problem due to the 900578392 tags entered
in the osm database as of last weeks planet (not counting relations).
If you're interested that's 12113 separate key values, or just 3087 if
you discount anything starting with "tiger". That makes interesting
stuff like what tags are used with what other tags, and queries such
as which way has a group of tags like xyz really difficult.

But until someone comes up with something that actually exists, the
best we have is the wiki and the hope that people just get on with it
and worry about tagging later :-)


PS. does anyone know where the 850,000 odd "name_1" tags comes from?
or the related name_2, name_3..... name_8 tags? That number makes them
more used than any other "name*" variant, but they're not mentioned
anywhere that I can find? I'm asuming Tiger.

More information about the dev mailing list