[OSM-talk] Semantics layer for tags

M∡rtin Koppenhoefer dieterdreist at gmail.com
Mon Jan 10 19:27:01 GMT 2011


2011/1/10 Martijn van Exel <m at rtijn.org>:
> (forgot to copy to talk)
>
> On Mon, Jan 10, 2011 at 6:35 PM, Ed Avis <eda at waniasset.com> wrote:
>> Martijn van Exel <m <at> rtijn.org> writes:
> The latter. The user would be able to tag a feature with "chemist",
> "pharmacy", "farmacia or "apotheek" and that would result in the same
> coding in the OSM database (currently: shop=chemist).


amenity=pharmacy, dispensing=yes/no

 When consuming
> OSM data, the process could be reversed; based on the locale, a
> feature tagged "shop=chemist" could (would) be output as being one of
> these culturally determined Things. Note that a "chemist", a
> "pharmacy", a "farmacia" and an "apotheek" are names for something
> that is similar across cultures and languages, but not literally the
> same.
>
> The idea is to *avoid* having different classifications on the
> database level, even though one concept could be represented by two
> different names in one language (consider freeway / highway). Any
> ambiguity arising from that would have to be handled by additional
> attributes.


I fear that a system like that will soon become utterly complex, thus
disabling most of the mappers of taking part in the
"tag-development-process". It would shift the discussions away from
the ML and wiki to defining the semantic rule set. And still we would
have to have definitions in natural language to define what a feature
is about, so there is no guarantee that there won't be contradictions
or different tags with the same meaning.

I agree that it is a good idea to develop such a ruleset (or extend an
existing one like linked geodata) to make the usage of our dataset
easier (for developers), but I agree with you: it is not a magic wand.

cheers,
Martin



More information about the talk mailing list