[OSM-dev] Street Addresses with Relations?

[Karl Newman]

On Dec 25, 2007 3:24 PM, Marcus Wolschon <Marcus at wolschon.biz> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Karl Newman schrieb:
> > Dave,
> >
> > I'm in active debate about this on the Wiki. The proposal page is a
> > mess, and I'm tempted to wipe it out and start over. Anyway, look at
> > my suggestion here:
> > http://wiki.openstreetmap.org/index.php/Proposed_features/House_numbers#Third_Suggestion_.28Using_relations.29
> >
> > My perspective on this is getting numbering to work on existing GPS
> > navigation devices (specifically Garmin). What I've proposed is
> > compatible with the requirements for that. I don't like the idea of
> > ranges because it could (will) lead to overlapping ranges, data
> > duplication at the endpoints, etc. Ranges just make it more
> > complicated than it needs to be.
>
> Dear Karl,
>
> I really like your proposal to tag only key-numbers at nodes and
> infer the rest (especially when ways are changed). Can anyone think
> of a way to optimize it for answering the question about the location
> of a single house-number without having to load all nodes of all ways
> that have a name similar to the one in the address?

I don't understand this question. Can you elaborate?

>
> However I think overlapping ranges can only happen in the same
> case where random house-numbers in your scheme will occur...if
> people are editing the road and moving nodes around and sometimes
> entering garbage (typo, swapped numbers, having >1 node selected,...).

No, overlapping numbers will occur if someone enters a range from the
first to the third node in a way, and someone else enters another
range for the second through the fourth (to give a simplified
example). This is not necessarily a problem until you start thinking
about the weird numbering cases, or the typos as you mentioned.
There's not really a solution for the typos, but their effects can be
minimized with my scheme (or at least easier to find and fix).

>
> A defined algorithm how to interpret the numbers and
> interpolate/extrapolate the missing numbers as well as examples for
> the usual cases would be a good thing.
> (linear-numbered sides, left-right -numbered sides, single house "2b"
>  in a regular road, road only containing house 12 and 18 but no houses
>  in between=how to disable interpolation.)

In that case you could just omit the street_number_type tag, which
makes it a "single" tag, with no implied range. But you wouldn't have
to--there are many cases where there are non-contiguous street numbers
(especially in rural areas with larger plots). I'd say just don't
worry about it and let the "phantom" street numbers exist.

> Just to prevent having house-numbers and every piece of software
> interpreting them in a different way.
>
> Marcus