[OSM-talk] House numbers... One more suggestion

Andy Robinson (blackadder-lists) ajrlists at googlemail.com
Tue Jul 29 12:06:38 BST 2008


Jan-Benedict Glaw wrote:
>Sent: 29 July 2008 11:33 AM
>To: Frederik Ramm
>Cc: talk
>Subject: Re: [OSM-talk] House numbers... One more suggestion
>
>On Tue, 2008-07-29 12:02:34 +0200, Frederik Ramm <frederik at remote.org>
>wrote:
>> Tagging house numbers onto intersections or ways is just a lame
>> workaround employed by many existing data sets because they can't
>> model every single house. I have absolutely zero doubt that sooner or
>> later, OSM *will* model every single house. At that time, every
>> single house will simply be tagged with its address and that's it.
>
>To me, it looks like a difference between modeling every house (and
>adding some kind of address tag) and describing house numbers along a
>way for eg. routing purposes.
>
>There are other schemes, which could work out quite well I guess:
>
>      119  121 ..    127     129   131 ...      137     139
>... x1 ------------------> x2 -----------------------> x3 ...
>      120   122  ..  128     130  132 ...       139     140
>
>House numbers are complicated (in a city like Berlin, you'll probably
>find non-continuous house numbers, and they not always follow the
>even-odd-scheme, and the house number on the opposite side of the
>street may even be in a completely other range...)
>
>x1: housenumber_righthand=120-128/+2
>    housenumber_lefthand=119-127/+2
>x2: housenumber_righthand=130-138/+2
>    housenumber_lefthand=129-137/+2
>x3: housenumber_righthand=140-148/+2
>    housenumber_lefthand=139-147/+2
>
>I only picked two tags and overloading them a bit, overloading them
>somewhat. It describes a starting house number and the common house
>number difference on the right/left side of the street. At least for
>Germany, that could describe house number distribution and the
>right/left information (which I find *quite* useful!) well enough even
>for the more exotic cases. However, it'll be somewhat of a pain
>optimizing a way.
>

The reason I don't like ideas like this is that the data you are adding to
the way (or as Frederik pointed out possibly also the intersections) is not
actually anything to do with the physical feature. There are no house
numbers or other references on a road, only the name of the street and
perhaps a road reference number. Accepted it's a nice easy way to make
routing work more easily but that doesn't make it right for our dataset. If
we keep and maintain the simple ideal that what we map is what we see then
it keeps it all very simple. Routing algorithms may need to be more complex
as a result but that doesn't give us an excuse to corrupt our data with
misleading information.

Cheers

Andy





More information about the talk mailing list