[OSM-talk] OSM relation ID property in Wikidata

[Christian Quest]

I'm not proposing to add this as a new tag to OSM objects, but to use
these geoURL in external datasets instead of OSM IDs.

It is a way to keep a (fuzzy) link between datasets.

Fuzzyness seems mandatory at some level as I doubt it is possible to
have an exact match between data models that are completely different.

With higher and higher level of details in OSM data, it becomes more
and more difficult to describe more global "things".
The node POI becomes a polygon POI, then maybe a multipolygon... the
single way street is later divided in several ways to describe each
part in details (pavement, oneway, light, trees, etc).

These geoURL could be generated from lat/lon + semantic infos about an
object, then resolved back to actual objects in whatever dataset
(including OSM but not only). overpass/XAPI could "resolve" these
geoURL by converting for example "shop at u09v8s9fd5" into
[shop=*][bbox=xxxx] as geohash are in fact bboxes.

Think of it like the internet DNS system... hostname can be resolved
to IP addresses and the reverse. You don't have to resolve hostname to
know you're talking of the same thing but you have to in order to
access it.


2013/5/7 Stefan Keller <sfkeller at gmail.com>:
> 2013/5/7 Christian Quest <cquest at openstreetmap.fr>:
>> Linking to OSM objet ID looks like a bad idea because they are not stable.
>
> Agreed.
>
>> I've proposed something that can be summarized like some generic
>> geoURL that could be used to describe something/somewhere in the form
>> something at somewhere
>
> Am I understanding you correctly: You propose a kind of geoURL - like
> linked data [1] - which is a tag attached to OSM nodes?
> That would fulfill most of my criterias except that it does not
> indicate that it's a "system property"
> GeoHashes have a nice property of being completely independent of a
> centralized (or partially centralized) id generator.
> But they are not suited to identify object which are overlaying each other.
>
> Yours, Stefan
>
> [1]  http://semanticweb.org/wiki/GeoURL
>
> 2013/5/7 Christian Quest <cquest at openstreetmap.fr>:
>> Linking to OSM objet ID looks like a bad idea because they are not stable.
>>
>> We had a recent long talk on talk-fr@ about permanent IDs that could
>> help linking external data to OSM, instead of multiplying ref:xxx in
>> OSM to link to external data.
>>
>> I've proposed something that can be summarized like some generic
>> geoURL that could be used to describe something/somewhere in the form
>> something at somewhere
>>
>> 'something' could use some hierarchical semantic description, and
>> 'somewhere' could use for example geohash.
>>
>> Example for a bakery nearby my place: saines_saveurs.bakery.shop at u09v8s9fd5mc
>>
>> These geoURL can be compared with some level of fuzzyness on something
>> and/or somewhere by reducing the level of details if necessary
>> (exemple: shop at u09v8s9fd5 has less details and can be matched with a
>> simple "contain" text search).
>>
>> They can be translated into whatever query like an overpass-api query
>> to actually resolve to the current matching objets in OSM or somewhere
>> else as the geoURL are not OSM based in any way.
>>
>> This mechanism works pretty well for node/polygon based POI, but needs
>> to be extended for more linear features (like a street), maybe using
>> more than one geohash (one at both ends ?).
>>
>> _______________________________________________
>> talk mailing list
>> talk at openstreetmap.org
>> http://lists.openstreetmap.org/listinfo/talk



-- 
Christian Quest - OpenStreetMap France
Synthèse du Week-end "SOTM-FR" à Lyon : http://openstreetmap.fr/synthese-sotmfr