[Imports] [Talk-cz] CzechAddress Import

Sarah Hoffmann lonvia at denofr.de
Thu Mar 13 06:52:32 UTC 2014


On Wed, Mar 12, 2014 at 11:28:01PM +0100, Petr Vejsada wrote:
> Dne St 12. března 2014 21:31:38, Martin Koppenhoefer napsal(a):
> 
> > This means you can add the name of the administrative district, but you do
> > not have to, it still remains a valid and unambiguous address. IMHO you
> > shouldn't tag this district to every housenumber. "Vinohrady" is redundant
> > information in this case and not necessary for the location to be found. Or
> > are there maybe also other "130 00 Praha 3 - xy" where xy is not
> > "Vinohrady"? Also for Boleslavska?
> 
> Aaaah, we are back again in the discussion about addr:place. It's 
> neverending...
> 
> You're right, in _this_ case, address without Vinohrady is valid. But:
> 
> - people knows, i'm living in Vinohrady. They usually don't know, i'm living 
> in Praha 3
> 
> - Nominatim really cann't find the house without tag addr:place. "Libochovany 
> 129" or "Vinohrady 1989" - nothing found. It was my primary motivation to run 
> this import - posibility to find what i'm searching for.

Alas, that won't change after the import, at least for this particular
address. For technical reasons, Nominatim can only support either a street
address or a conscription number. So, it will work fine in small villages
that have only conscription numbers but not in Prague where there are both.
But that is a missing feature in Nominatim.
I believe that your tagging in this case is completely correct. The moment
you add a addr:conscription_number tag, there should be an addr:place tag
with the name of the place the number belongs to. Just like there should
be an addr:street tag when addr:street_number is added. It has exactly the
same function: relate the number to the object that determines the address.

> There is some redundancy, but i don't think enormous ...
> 
> There is polygon of country boundaries - addr:country=CZ is redundant. We 
> explained several times, why to use this tag Yes, it's about geocoders. It's a 
> part of address.
> 
> There are polygons of city boundaries - addr:city=* is redundant. City is part 
> of address.
> 
> There are NO polygons of "cast obce", inexistant boundaries we can't import 
> into OSM - addr:place=* is not redundant. addr:place is not cadastral place. 
> addr:place is nececessary part of address, if there is no 
> addr:suburb/borough/...

addr:suburb/borough is not the same as addr:place. Please don't use one in
place of the other. Always add addr:place. addr:suburb/borough is redundant
in the same way addr:city is. Most of the time it can be determined from
the boundaries. It is only really useful where the postal address differs
from the admin boundaries. (Happens more often than one would like.)

If you still want to add it, addr:suburb should be sufficient, no need
to invent a new tag. I'd interpret 'suburb' here just as: the part of
the address that describes a part of the city. And I believe that fits.

> Agree, it's a little bit diferent. This is amenity point, not address point. 
> Restaurant is in house (sometimes, garden restaurant is not) and the house 
> usually has their's own address point.
> 
> > I don't even need to mention what borough (or county) the object is in
> > for the same reasons.
> 
> > Adding in the other fields is just a matter for a geocoder.

That argument doesn't really work. Every geocoder needs to be able to
process boundaries because 95% of the addresses don't have a complete
set of addr:* tags. And once you have boundary support, adding additional
support for the addr:* tags doesn't really add much. Incidentally, that
is the reason why Nominatim still sucks at processing most address tags.

Sarah



More information about the Imports mailing list