[OSM-talk] is_in

[David Earl]

On talk-gb, Shaun McDonald [mailto:shaunmcdonald131 at gmail.com] said of the
name finder:

> 3. Does the is_in tag work recursively? That way I could simply say
> is_in=Edinburgh, and it would then automatically get all the is_in
> tags from the nearest Edinburgh place. Edinburgh would then get all
> the is_in tags from its nearest is_in tags. This would save a lot of
> duplication in the database, while easing data entry.

This isn't actually related to name finder as it doesn't use is_in at all to
do its searches (it is done by proximity).

But it does raise difficulties with is_in that have troubled me.

HIERARCHY

The principle of hierarchy is a good one IMO, if it is indeed hierarchical
(see below). I could use it very effectively in name finder, and it would
avoid growing inconsistencies.

But it isn't clear where one should start. At the moment most creators of
is_in seem to put the town or city of which a suburb is part, or the county
for a village or town.

We could argue though that in the UK one should include the district if
different:
  Fulbourn is_in=South Cambridgeshire,Cambridge,England,UK

Ideally we'd just have
  Fulbourn is_in=South Cambridgeshire
  South Cambridgeshire is_in=Cambridgeshire
etc and let the system work it out (I could certainly do so in the name
finder if it was like this).

But ideally the information shouldn't be needed at all, if we had
administrative boundaries represented. Point in polygon tests would then
enable this. But getting administrative boundaries in is hard (we can't
trace them from modern maps). Actually the other way round might be
interesting - estimate the bounaries by examining is_in, which wouldn't
establish the boundary, but at low zooms could produce a useful map.

Is is_in actually hierarchical anyway? Are district level authority
boundaries always enclosed within regional ones (District, County in the UK
and whatever equivalents there are elsewhere). For example, currently the
South Cambridgeshire parliamentary constituency overlaps both Cambridge and
South Cambridgeshire districts but encloses neither completely. (Not that
I'm suggesting is_in should record parliamentary consituencies).

LANGUAGE

What language is is_in in? Some people have used English even where it isn't
natural (...,Germany) rather than (...,Deutschland), others have used the
native language.

My preference would be to use the local language, and determine alternates
(which needn't be English of course) from the place tag for that entity
(possibly hierarchically).

David