[Tagging] Places & admin boundaries

[Martin Koppenhoefer]

2012/10/28 Frederik Ramm <frederik at remote.org>:
> On 27.10.2012 17:53, "Petr Morávek [Xificurk]" wrote:
>> 2) Boundary relation or place=* for higher administrative units?
>> The overlap is especially visible in case of countries (but same
>> arguments apply for other units like states, counties, ...).
>> Majority of counties are in database represented twice - as node with
>> place=country and with admin_level=1 boundary relation.
>
>
> Note however that there is no 1:1 relationship between place=... and
> admin_level - e.g. something on admin_level 8 could be a place=city or
> place=town, or (in Germany) something on admin_level can be a place=state or
> place=city capital=yes or so.


+1, there could also be several places (e.g. villages or towns) within
the same admin_level-8 entity. Place is a more open and less defined
concept, geographical places (toponyms), who depend on different
cultural factors, while administrative division is a clear and well
defined order (at least usually there are clearly defined and codified
borders), besides some disputed areas where things are not that clear
in this layer either.

IMHO places could on the longer run be approximated by areas as well,
even if their borders are not as clear as they are for administrative
boundaries (it's a wiki and it will evolve to some compromise that is
acceptable for most). For "routing" you don't really need a point in
the center, normally you would have a more precise address where you
want to go, and in the absence you could just as well use an arbitrary
node to route to (e.g. centroid, or the label-node position), for
rendering it is also not strictly necessary, but if you wanted to have
a node for the label why not simply tag it like this? (place-label).
In the end a place is never a "point", and if you nonetheless need one
for your graph-model, you can take any arbitrary point that makes
sense ;-)

cheers,
Martin