[OSM-talk] Part-municipalities and more place issues

[Ben Laenen]

On Sunday 30 December 2007, Martijn van Oosterhout wrote:
> TBH I think the word suburb fits perfectly. Suburbs in AU can have
> any number of residents. It's mostly a division made for the purpose
> of streetnames and postcode boundaries. To what level is streetname
> uniqueness: to the "deelgemeente" or 'gemeente"?

In general the street name uniqueness is for deelgemeentes.

Well, suburb could fit as a term for districts of Antwerp, but I can't 
really call "Knokke" or "Heist" a suburb in "Knokke-Heist". And calling 
a settlement with a few dozen inhabitants several kilometers away from 
the main settlement in that municipality which has thousand inhabitants 
a suburb also stretches its meaning a lot...

The problem we do have with using suburb is that there are really 
smaller "quarters" inside each deelgemeente/district which also need 
tagging, and I'm using suburb for those now... (or could "hamlet" be 
used for that? I thought hamlets were just very small settlements far 
outside any village)


> > * To be quite blunt: I don't like the usage of the city/town/...
> > tags: I'd like it much more that there are general tags like
> > "municipality" (and in our case also something like
> > "part-municipality"), and use an extra "population" tag to know how
> > big to render the name. In Belgium "city" is a title given to a
> > municipality (I think the UK has the same system), and if we tagged
> > every city as such over a fifth of all municipalities would be a
> > city (133 of 589), which would make the map of Belgium quite
> > interesting...
>
> Maybe you shouldn't give them all the same tag? Maybe you should use
> city/town/etc *and* use the population tag. Best of both worlds.
> Also, you problem of how the map would look is interesting. If each
> city had a population tag the mapnik could display them all nicely
> showing the biggest first and show smaller ones as there is room.
> Sounds good to me (how osmarender wold deal with this I have no
> idea).

Okay, I was rather joking about giving them all city tags. I was just 
trying to point out that the term isn't really useful, and can give 
confusion where these terms have specific definitions in a country.

> Please not another tag..... "population" already exists as does
> place=town/city/foo. Let's use those.

Sure, but if it exists, why not use it then to render it on maps? And if 
we can have that, why do we still need all the different 
city/town/village tags to distinguish them (other than for those where 
no population is available)?

> > The way I'm personally tagging things in Belgium now is: discard
> > all info about municipalities/districts, and use for example "town"
> > for a district of 70.000 inhabitants. Surely that's a nice way to
> > get it rendered nicely, but you'd lose the data that one belongs to
> > the other one. is_in wouldn't do its job to fix that:
> > is_in=Antwerpen could mean anything from province, arrondissement
> > (group of municipalities, we need a tag once for that too maybe,
> > but that's not urgent, we don't really use that one :-) ),
> > municipality and district...
>
> For is_in you should try using relations. Then you have what contains
> what without haveing to mess with any tags at all...

I haven't investigated is_in relations, but does that mean that you put 
two nodes into one relation a bit like:

member:          role:
deelgemeente A   -
deelgemeente B   -
deelgemeente C   -
gemeente D       belongs_to

(which would give difficulties I guess if you need to put a town in a 
country, where to get that country node to assign to relation to, can't 
really download entire countries in JOSM :-) ... I need to read about 
it on the wiki once :-) )

Hmm, looking at the wiki it doesn't have any "belongs_to" role... Not 
entirely sure that can work well (read: give conflicts if bigger and 
smaller places have the same name).

> > Anyway, I still need to find a nice way to render the
> > Houthalen-Helchteren case, so it's not like:
> >    Houthalen
> >   Houthalen-Helchteren
> >            Helchteren
> > on the map.
>
> Rendering is another problem entirely and what you want isn't really
> possible with any current renderers. Perhaps we should wait to see
> what the renderers could possibly work with before deciding on what
> tags might be useful. Instead, focus on what we do know: getting
> accurate data in the DB.

Well, I have the habit of thinking thoroughly and testing out the way 
I'm tagging things. We need to agree on something between the Belgian 
mappers, as right now everyone is doing as they see fit. And at the 
moment it seems to be a conflict between getting it to render correctly 
and getting in the exact tree of dependencies right, so we're looking 
for other opinions before making the conventions.

So, at the moment I guess we need something like:
* Don't look at city/village/town/suburb tags if there is a population 
tag available to render things on the map (okay, not entirely necessary 
if we do the following point)
* Keep tagging the part-municipalities like town/village with the usual 
agreements (> 10.000 = town etc.) (by discarding the fact that they 
belong to municipalities)
* Use is_in relations to declare which towns/villages belong to which 
towns/villages/cities

Anyway, does that look wrong to anyone?

The only problem I see with this is that the part-municipality which has 
the same name as its municipality will render both at the same time. 
It's not really pretty, but I can live with that. (Hmm, but it would 
mean two "city" nodes for Antwerp, can a node be "place=city;town" and 
can a node belong two times to a relation with different roles?)

> Have a nice day,

Greetings
Ben