Hash: SHA1

Am 17.07.2014 18:17, schrieb Robert Sanderson:

> If both URIs identify the same "thing", in this case a real world person,
> then IMO the correct predicate is owl:sameAs, and there is no need for a
> bf:Identifier.
> For people, where there isn't any need for interpretation, this is
> relatively easy.  Either a URI identifies me, or it doesn't.  For more

ISNI and ORCID are both about "persons". However ISNI has granularity
at the level of "bibliographic identity" and ORCID is more like "real
individuals". Bibliographic authority files typically lie in between,
but also extend their realm to deities, fictitius characters and so
on. Of course for 99% of the resources described this does not matter,
but for the remaining cases less apodictical statements than "owl:sameAs"
seem appropriate.

> resources that require interpretation, such as bf:Work, or Topic, the
> same-ness is much trickier to determine as the exact nature of the resource
> is difficult (I might even say impossible) to determine.  Thus whether an
> ISBN identifies exactly the same "thing" as another similar identifier is
> much trickier.

ISBNs establish a relation with respect to certain aspects of the "thing"
as defined by the ISBN agency. Thus a classical MARC21 AACR2 record with
its mix of work level and manifestation level data elements and a
bf:instance graph both may legitimately carry the same ISBN as an identifier.
Whereas stating them to be owl:sameAs demands prior reinterpretation of
the MARC record in terms of FRBR.

>> Now consider
>> <> a bf:person;
>>    bf:identifier [
>>        bf:schema "VIAF";
>>        bf:identifierValue "195531823";
>>        bf:identifierValueURI <>
>>      ].
> Or ... <> owl:sameAs <
>> .

... loosing the tight association of the identifierValue "195531823"
with <> or ""

>> However identity with the resource <>
>> is NOT built into that construction, it's rather "all resources
>> relating the same way with <>
>> are pairwise equivalent" and someone would /explicitly/ have to add
>> <> a bf:person;
>>    bf:identifier [
>>        bf:schema "VIAF";
>>        bf:identifierValueURI <>;
>>        bf:identifierValueURI <>;
>>        bf:identifierValueURI <|n89613425>;
>>      ].
> And here is demonstrated the issue with approaches like this...
> X a class ;
>   bf:identifier Y .
> Y a bf:Identifier ;
>   bf:identifierValueURI X .
> is unnecessary circularity.  Also, having multiple bf:Identifiers, or
> multiple values on a single blank node like above introduces additional
> tiers of semantic complexity.

I consider bf:identifierValueURI <>
as sort of an "opaque token" with URI (or rather IRI) semantics
when it comes to comparisons.

Of course, X is a resource and when I formulate statements with
X in subject position then these X must be the same resource.

On the other hand you'll never have tautological statements like
X bf:identifierValueURI X
and you could consider the catenation of the relations
bf:identifier o bf:identifierValueURI
as kind of split of the owl:sameAs direct relation.

So although there is one extreme case with a tautological outcome
the concept of bf:Identifier in the general case gives us the
headroom necessary to deal with any representation of identifiers.
Thus for me still no argument that bf:identifierValueURI is
definitely a bad thing to do and rather bf:identifierValueURIasString
should be the way to go.

viele Gruesse
Thomas Berger
Version: GnuPG v1
Comment: Using GnuPG with Thunderbird -