[Tagging] part_of:wikidata key

[Kevin Kenny]

On Tue, Nov 28, 2017 at 9:01 AM, Christoph Hormann <osm at imagico.de> wrote:
> On Tuesday 28 November 2017, Andy Mabbett wrote:
>> > the established rule not to have external IDs in the
>> > OSM database
>>
>> What "established rule" would that be? Established by whom; when and
>> how?
>
> I am sorry for the ambiguous wording - as Fred said i meant essentially
> a custom - calling it "established" was meant to indicate that.
>
> You could however argue that this stems from the evidently existing rule
> of on-the-ground verifiability.

Of course, there's room for judgment. We routinely add wikipedia=*
or url=* tags to say, "look here for more (non-map) information about this
map feature." Of course, it often isn't field-observable that a given facility
has a web site, or that a noteworthy geographic feature has a Wikipedia
article, but the link is useful information, and we put it in.

We will add contact information for a facility even when not all the
information is on the sign, for the same reason: it's public, it's useful,
it's reasonably stable.

I tend to feel the same way about *certain* external database keys.
NHD reachcodes, for instance, are documented to be stable, and used
by other hydrographic databases as identifiers for watercourses.
National Register of Historic Places registration numbers are also
persistent, and usable to key into a lot of information about the places,
since the registration packets contain lenghty discussions about their
significance.

Until this discussion, I thought that wikidata=* tags were a similar
informational aid: "here is an external resource that has more
information about this feature." Mostly harmless.

What troubles me in this discussion is not that wikidata is an external
database key. Rather, it's that I appear to be hearing an expectation
that its presence implies an expectation that OSM will conform with
an expectation of some external database that the object tagged with
wikidata must be unique and that its identification is stable. That's
perhaps conceivable for point features, or for relations created for
the purpose. It's surely not feasible for ways. I've split ways many
times to correct topology, to put a route relation, traffic restriction,
or other attribute on part of the way, or even because the way simply
visited too many nodes to be convenient to edit. I surely wouldn't
have known that it's not permissible to do so with 'wikidata' tags,
and I find it astonishing that anyone thought that was reasonable.
For multiple objects to assert the same telephone number or
website is harmless and routine. Nowhere is there an implication
that a telephone number will be bound to a unique OSM object.

I could be convinced that creating relations to hold wikidata
is a good idea, but I would need to understand the use cases and
what value they add to the map. Wikidata, to me, is mysterious.
I've read up on how it's essentially a key-value store for noteworthy
subjects, with links to sites such as Wikipedia, OSM, a subject's
own web site, service directories, and so on, but I don't believe
that I've ever actually seen an application that uses it, and don't
have a real understanding of how it is used and what can be done
with it that cannot be done without it.