Print

Print


Dear Rob,

A few quick responses to your email followed by a longer excursion, unrelated to points you explicitly raise but still about identifiers in bibframe.

> An example in the documentation of when this isn't the case is in bf:uri
> (  http://bibframe.org/vocab/uri.html ) where there is a real resource as the
> object of the bf:uri predicate.
> 
> The example asserts, via the inference from the range, that:
> 
>   _:bnode1 a bf:Instance ;
>             bf:uri <http://lcweb2.loc.gov/pp/cphquery.html> .
>   <http://lcweb2.loc.gov/pp/cphquery.html> a bf:Identifier .
> 
> And thus that it could have properties like identifierQualifier,
> identifierAssigner and so forth.  Maybe that's the intent, but it doesn't really
> seem that way?

I think this is a mistake.  So, actually, thanks for bringing that to our attention.

> Now consider:
> 
>   _:LotR a bf:Instance ;
>       bf:uri <http://example.com/identifiers/book1> .
>   <http://example.com/identifiers/book1> a bf:Identifier ;
>       bf:identifierValue "some string identifier here" .
> 
> Now it becomes very odd. 

Oh, yes, it is extremely odd.  The more so because every time I look at that example I think to myself, "who in their right mind would do that?"  :)   Actually, my confusion comes a little later when the URI for the bf:Identifier is suddenly the same as that for the bf:Instance. Please don't misunderstand, I comprehend your overall point, but I remain confused as to why anyone would use the URI of an Instance as the URI for a bf:Identifier.  It seems like that individual is just asking for trouble.

Anyways, in the end....

> The solution (as above) is just to use bf:Identifier, always as a blank node, for
> only identifiers that are not themselves URIs.

I think I can sign on to this, but I'm not confident that this rule resolves this particular identifier issue as I see it.  And so begins the excursion...

After reading, and re-reading, and re-reading your email and document, I can't help but think the problem is the bf:uri property itself.  I've had reservations about this particularly identifier property (and few other ones in the vocabulary) and I wonder if the trouble you are seeing isn't in fact a slightly different manifestation of the problem I see.  Which is:  What does property bf:uri seek to do?

It's a property, so it should relate two things, yet I do not believe the right semantics are being captured in this case.  I agree, if you have a URI, you have a URI.  It identifies something already.  It's the "what" that I do not think is being addressed.  Is the URI an identifier for the current resource?  That is, is it about same-ness (some form of equivalency or owl:sameAs relationship)?  Or, given a different context, the thing you have might be a URI but the appropriate relationship is something else altogether.  Of course, you'd want to use a property that captures the nature of the relationship versus the rather non-descriptive bf:uri.  To ask all of that another way, when does one use bf:uri?

Does this

 _:LotR a bf:Instance ;
     bf:uri <http://example.com/identifiers/book1> .

mean this?

 _:LotR a bf:Instance ;
     bf:equivalentInstance <http://example.com/identifiers/book1> .

(I might have made up the bf property for demonstration purposes, but you get the idea.)  If so, then would it not be better to use bf:equivalentInstance, which is far more explanatory, than bf:uri?  Or is knowing the type of identifier more important?  (Really, those are not rhetorical questions.)

bf:hdl (http://bibframe.org/vocab/hdl) is another one that I find similarly lacking in strong semantics.  Add bf:urn to that as well.

This last case is interesting because a URN is a URI.  So, to be clear, would bf:urn fall under the rule your proposed above?  (Your document would suggest as much, which is all fine by the way I'm just looking for clarification.)

The bf:identifier property has a number of sub-properties (http://bibframe.org/vocab/identifier).  Nearly all of those derive in some way from their existence in the MARC 01X-09X fields. Many in that list derive more specifically from the MARC Identifiers source list, the values from which are expected to be used in a subfield of the MARC Bib 024 (Other standard identifier).  The 024 is an identifier for the resource described by the MARC record; the identifier "code" used in the subfield tells you the type of identifier.  In BF, these identifiers have been published as properties (relating a Work to a bf:Identifier resource or an Instance to a bf:identifier resource), but does that capture the intended semantics?  

Does this capture the semantics better

_:LotR a bf:Instance ;
    bf:doi _:bnode1

_:bnode1 a bf:Identifier
   bf:identifierValue "10.1000/182"
   bf:identifierAssigner <http://example.org/1>

than this

_:LotR a bf:Instance ;
    bf:identifier _:bnode1

_:bnode1 a bf:Identifier
   bf:identifierType <http://id.loc.gov/vocabulary/identifiers/doi>
   bf:identifierValue "10.1000/182"
   bf:identifierAssigner <http://example.org/1>

?

(And, for the record, we can of course present DOIs as HTTP URIs, but let's work on this example here.)

Perhaps it makes little difference and I'm splitting hairs.

That said, one of the points you raised in the document was how, for example, an ISSN has a formal URN form and therefore could be represented as a URI (did not know that, so thanks).  Taking advantage of such a mechanism, I could see a scenario where the property bf:issn is defined as something like "ISSN for current resource in URN form" and reserved solely to allow something like this:

_:I1 a bf:Instance ;
    bf:issn <urn:ISSN:1560-1560> .

while still allowing this:

_:I1 a bf:Instance ;
    bf:identifier _:bnode1

_:bnode1 a bf:Identifier
   bf:identifierType <http://id.loc.gov/vocabulary/identifiers/issn>
   bf:identifierValue "1560-1560"
   bf:identifierAssigner <http://example.org/issn/agency>

The immediately preceding two examples would establish a model whereby "bf:identifier" and "bf:Identifier" are used when capturing information about other, string-based identifiers while reserving the properties (bf:issn, bf:isbn10, etc) exclusively for URI-ified identifiers.  This aligns, I believe, more or less with your rule; I'm wondering if there is a way to refine when select properties are used and when they are not (providing there is merit to that idea).  I'm also looking at how they are defined.  Does that make sense?  Do you foresee issues with or benefits of allowing for this type of duplication?

OK, I feel I rambled on a bit; I hope it was clear.  

All of it is to basically say that I think you raise a good point, but that I am trying to flesh it out some.  Thanks for taking the time to write it all down.

Yours,
Kevin

--
Kevin Ford
Network Development and MARC Standards Office
Library of Congress
Washington, DC


> -----Original Message-----
> From: Bibliographic Framework Transition Initiative Forum
> [mailto:[log in to unmask]] On Behalf Of Robert Sanderson
> Sent: Wednesday, July 09, 2014 7:15 PM
> To: [log in to unmask]
> Subject: [BIBFRAME] BibFrame and Linked Data: Identifiers
> 
> 
> 
> * Identifiers
> We do believe that identifiers in BIBFRAME could benefit from more thinking.
> As such, we feel that Rob’s section on identifiers may be trying to grapple
> with several different issues, all of which are about identifiers but each of
> which should be treated separately. These issues are likely legitimate, but we
> need to sort them out.
> 
> Agreed, this is a good observation, and I'll try to tease apart some of those
> issues below.
> 
> We do want to note, bf:identifierValue expects a string literal, and Rob’s
> examples appear to treat it as a resource identifier.  This is a fairly
> fundamental misinterpretation that needs to be cleared up before we can
> begin to sort out the other issues.
> 
> This wasn't my interpretation. The documentation is very good about when a
> literal or a resource is expected as the range of a predicate, kudos on that!
> 
> Let me start by saying I completely understand and agree with the existence
> of bf:Identifier.  It is important to be able to capture *non-URI* identifiers
> for resources, especially in such a way as to record qualifiers, assigners and
> schemes/namespaces that they might fit into.
> 
> * Point 1: Identifier Unnecessary for URIs
> 
> It's when the identifier is a URI that things become problematic.  If there is a
> resource that has an identifier which is a URI, then in RDF that URI is the
> identifier for the resource.
> Thus:
>   _:bnode1 a bf:Instance ;
>      bf:uri _:bnode2 .
>   _:bnode2 a bf:Identifier ;
>      bf:identifierValue "http://www.example.com/books/book1" .
> 
> Is really saying, at least to me:
>   <http://www.example.com/books/book1> a bf:Instance .
> The first reads as: "There is a resource without an identifier, an Instance, and
> it has an identifier that's a URI." Clearly, this is syntactically possible to say,
> but is self-contradictory.
> Or am I mistaken? If I am, then I think there needs to be some very thorough
> documentation about the difference between the Bibframe notion of an
> identifier and the RDF/Linked Data notion of an identifier.
> 
> * Point 2:  Identifier Must be a Blank Node
> 
> The key point here is that bf:Identifier must always be a blank node.  Why am
> I saying this?
> 
> If the bf:Identifier gets given its own identifier, then there are some very
> strange semantics going on. Consider the following:
> 
>   _:LotR a bf:Instance ;
>       bf:uri _:id .
>   _:id a bf:Identifier ;
>       bf:identifierValue "some string identifier here" .
> 
> No problem, right? There's an Instance which has a string associated with it,
> which identifies the Instance. Right! All is good.
> 
> Now consider:
> 
>   _:LotR a bf:Instance ;
>       bf:uri <http://example.com/identifiers/book1> .
>   <http://example.com/identifiers/book1> a bf:Identifier ;
>       bf:identifierValue "some string identifier here" .
> 
> Now it becomes very odd. This asserts that there is some resource,
> (identifiers/book1) which has an identifierValue. But it's not really that
> resource's identifier, it's the Instance's identifierValue that we want to assert
> other information about (like the assigner, qualifier, etc)
> 
> It would be solved by being clear that bf:Identifier is only to be used for
> identifiers of the resource that are not URIs, AND that the construction must
> be a blank node.
> 
> * Point 3:  Identifier Collision for URIs
> 
> In the situation where there's a uri associated with an Instance, such as:
> 
>   _:bnode1 a bf:Instance ;
>      bf:uri _:bnode2 .
>   _:bnode2 a bf:Identifier ;
>      bf:identifierValue "http://www.example.com/books/book1" .
> 
> It seems like the Identifier should be identified by the URI
> <http://www.example.com/books/book1>.  If there's something which that
> URI identifies, then how can it not be itself?
> 
> Thus we get:
> 
>   _:bnode1 a bf:Instance ;
>      bf:uri <http://www.example.com/books/book1> .
>   <http://www.example.com/books/book1> a bf:Identifier ;
>      bf:identifierValue "http://www.example.com/books/book1" .
> 
> Which seems redundant but not terrible.  However, if we know the URI that
> identifies the instance, we should use that URI instead of a blank node. Yes?
> 
>   <http://www.example.com/books/book1> a bf:Instance ;
>      bf:uri <http://www.example.com/books/book1> .
>   <http://www.example.com/books/book1> a bf:Identifier ;
>      bf:identifierValue "http://www.example.com/books/book1"
> 
> Which is actually:
> 
>   <http://www.example.com/books/book1> a bf:Instance, bf:Identifier ;
>      bf:identifierValue "http://www.example.com/books/book1" ;
>      bf:uri <http://www.example.com/books/book1> .
> 
> Now we have a mess! :(
> 
> 
> The solution (as above) is just to use bf:Identifier, always as a blank node, for
> only identifiers that are not themselves URIs.
> 
> An example in the documentation of when this isn't the case is in bf:uri
> (  http://bibframe.org/vocab/uri.html ) where there is a real resource as the
> object of the bf:uri predicate.
> 
> The example asserts, via the inference from the range, that:
> 
>   _:bnode1 a bf:Instance ;
>             bf:uri <http://lcweb2.loc.gov/pp/cphquery.html> .
>   <http://lcweb2.loc.gov/pp/cphquery.html> a bf:Identifier .
> 
> And thus that it could have properties like identifierQualifier,
> identifierAssigner and so forth.  Maybe that's the intent, but it doesn't really
> seem that way?
> 
> Thanks for thinking about these, and also helping me to think about the
> separate points :)
> 
> Rob
> 
> --
> Rob Sanderson
> Technology Collaboration Facilitator
> Digital Library Systems and Services
> Stanford, CA 94305