[OSM-talk] Hierarchy of places

[David Earl]

The discussion of finding Paris Ontario equated to Paris France just now 
reminds me to raise again the granularity of our place hierarchy. 
Notwithstanding value judgements people make about what is a "town", not 
just based on population, I still think we need some more levels in the
city - town - village - hamlet hierarchy.

I could use population in the namefinder (and ditto for caption size, 
priority etc in renderers) if that were given more widely, but the 
problem is that in the UK at least this information is subject to crown 
copyright (I assume any one figure is a fact, and therefore not 
copyright, but the collection is database copyright).

I think we could do with a richer hierarchy something like this:
metropolis    > 500,000
city          > 100,000
large_town    > 25,000? 40,000?
town          > 10,000
small_town / large_village > 2,500
village       > 100
hamlet        < 100

I'm sure we can argue about the names and numbers, but I don't think 
that's too important as people will always bring their own local 
knowledge to bear as well, not least based on what the place chooses to 
call itself. For example, Hay-on-Wye (Y-Gelli), Powys is most definitely 
a town even though it has fewer than 2,000 people, but Linton, 
Cambridgeshire is a "village" of nearly 5,000 souls.

Another possibility is to make judgements about "importance" based on 
land area. I think this would be hard, though not completely impossible, 
to infer automatically from the data, but I wonder if a "radius" value 
on the place might help - "a circle approximately this big would enclose 
the place". Perhaps over-exaggerates coastal settlements and others 
which aren't blobs, so can only be a coarse judgement, but it might help 
me in deciding what "in" and "near" mean in the context of searches, 
especially for places which "punch above their weight".

David