[Tagging] "part:wikidata=*" tag proposal for multiple elements connected to the same wikidata id

[Paul Allen]

On Thu, 12 Sep 2019 at 09:45, Janko Mihelić <janjko at gmail.com> wrote:

One problem with the current system is that if you click one of those
> dwarfs in OSM, and see it's linked to an object in wikidata, you have no
> way of seeing if that is the whole wikidata object, or just a part of that
> object, unless you download the whole OSM database. Or if you are a human,
> and you look at
>
 the wikipedia article, and see there should be a whole bunch of dwarfs.
> But that example doesn't seem as important.
>

You're trying to solve the data problem (to the extent that it even is a
problem) in a defective
way.  It's defective because wikidata is not an integrated part of OSM
designed to group objects.
It's outside of our control; even if we can abuse it that way,  a random
edit to wikidata will
lose the object grouping.  The correct way to group them is with a
relation.  If we don't have a
suitable type of  relation then propose one.  Don't use wikidata as a
workaround for not having a
suitable relation type or not having a part_of_a_group=yes tag.  And even
without any of that,
an overpass query in the general area for artwork_type=statue + other tags
they have in
common will find them.

It's not a problem for humans either.  If they're not interested in looking
at the data item then
it doesn't matter if it's tagged wikidata=* or part:wikidata=* because they
won't follow it.  If
they are interested in looking at the data item then it doesn't matter if
it's tagged wikidata=*
or part:wikidata=* because they'll reach the same data item either way and
realize there
are seven dwarfs.  And won't be able to find the other six easily from
there.  So a relation
is still the best way to do it, then apply wikidata=* to the relation.

Currently, the second most numerous wikidata tag in OSM is
> https://www.wikidata.org/wiki/Q2961670, an item that describes all the
> roman roads in historic Gaul in France. All those ways, close to 500 of
> them, have wikidata=Q296167. That is obviously not good tagging. But how do
> you differentiate good wikidata tagging from bad tagging? I think this rule
> and part:wikidata are the way to clean this up. I would give all these
> roads part:wikidata=Q29616, and than that looks much closer to reality.
>

I think the only sensible solution is to delete the wikidata tags from
*all* of them.  That item is
for a category, not a unique object.  OSM relations are not categories
because we don't tag
categories (it would result in a gigantic taxonomic hierarchy of tagging).
Those road should
never have been given that wikidata tag; individual roads get a wikidata
tag only if that data
item applies solely to that particular road.

It seems to me that you're trying to find a way of mapping anything that
has a wikidata tag.
If so, that seems like a bad idea.  Use a wikidata tag to add extra
information about a unique
OSM object, don't invent OSM objects and/or ways of mapping things in order
to put every
wikidata item into OSM.  We could put wikidata=Q2 on every object,
following your line of
reasoning.  Oh, sorry, part:wikidata=Q2.

-- 
Paul
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tagging/attachments/20190913/f98ae2d4/attachment.html>