[Tagging] Part/whole confusion with Wikidata tag, and the need for enveloping parts into a whole
yuriastrakhan at gmail.com
Tue Aug 7 13:43:58 UTC 2018
Christoph, you are right that some loosely defined areas like rainforests
may not have exact boundaries. We can find limitations and issues in
defining/naming/linking pretty much anything, e.g. see discussion for
"[Tagging] place nodes for continents". That said, in a large number of
cases, it is beneficial to data consumers to have a 1:1 mapping, e.g. for
examples presented by Peter and François. We do not have to extend that
approach to objects that it won't work well for.
So if it makes sense, it is ok to have a "concept-level" relation that
defines common properties, such as shared wikidata ID, perhaps a relevant
Unesco Heritage ID, a URL, or the hours of operation. It would be a bit
silly to repeat that info on every part of the location.
And for other types of objects, especially the ones without a clear
outline, perhaps it may not make sense to even add wikidata IDs at all.
On Tue, Aug 7, 2018 at 11:04 AM Christoph Hormann <osm at imagico.de> wrote:
> On Monday 06 August 2018, peterkrauss wrote:
> > Seems a commom quality problem of part/whole confusion in the
> > Wikidata attribution or OSM's POI reference... And where there are a
> > need for "enveloping parts into a whole".
> > [...]
> The fact that there is no agreement on the nature of the relationship
> between Wikidata objects and OSM objects has been an important point of
> critique of the whole 'adding wikidata IDs to OSM' movement. You can
> read this up in the previous discussion here and in talk.
> OSM aims to map based on local verifiability. Therefore many things we
> map in OSM have no equivalent in Wikidata (because they do not satisfy
> the criteria for inclusion there) and many things in Wikidata cannot be
> mapped verifiably in OSM. And inventing some kind of collector
> relation that collects all objects that by some wikidata
> interpretation 'belong to' a certain Wikidata ID and thereby implements
> a 1:n relationship would not change that (it would just be pointless
> non-maintainable, non-verifiable dead weight in the database).
> My favourite example for this is the Amazon rainforest (but you can use
> other large eco-regions like the Sahara desert as well). You won't be
> able to verifiably map the Amazon rain forest in OSM as an entity.
> What we aim to do in OSM is to accurately map the woodlands of South
> America - which is still a very long way to go. But if this should
> happen it will happen locally because natural=wood/landuse=forest is
> locally verifiable while the abstract concept of naming some of this
> woodland the Amazon rainforest is not.
> Christoph Hormann
> Tagging mailing list
> Tagging at openstreetmap.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Tagging