[open-bibliography] Identifiers

Mon Nov 22 16:19:04 UTC 2010

William,

I think it would be best if DC upgraded their vocabularies to OWL to avoid this ambiguity. Here is a suggested start for the overhaul:

:identifier a owl:DatatypeProperty ;
	rdfs:subPropertyOf dcterms:identifier .

:identifierRef a owl:ObjectProperty ;
	rdfs:subPropertyOf dcterms:identifier .

OWL is mentioned a few times in the recent DcamInContext document, but only in relation to application profiles and constraint languages:

http://dublincore.org/architecturewiki/DcamInContext

I think that OWL should be more fundamental. For example, the DCMI Type vocabulary looks like it should be a plain old OWL ontology suitable for use with rdf:type:

http://purl.org/dc/dcmitype/
http://dublincore.org/documents/dcmi-type-vocabulary/

Jeff

> -----Original Message-----
> From: public-xg-lld-request at w3.org [mailto:public-xg-lld-
> request at w3.org] On Behalf Of William Waites
> Sent: Monday, November 22, 2010 8:40 AM
> To: List for Working Group on Open Bibliographic Data
> Cc: public-xg-lld at w3.org
> Subject: Identifiers
> 
> (cross-posting a bit to the LLD XG to benefit from the insight of
> people there that may not be subscribed to open-bibliography@)
> 
> * [2010-11-22 13:49:46 +0100] Makx Dekkers <mail at makxdekkers.com>
> écrit:
> 
> ] I am not sure what you mean by "confusing in that it was a literal
> that
> ] was constructed like a URI".
> 
> This is not a question of right or wrong, but a question of best
> practices.
> 
> ] In the sample I see:
> ]
> ]     <dcterms:identifier>GBA164362</dcterms:identifier>
> ]
> ] and
> ]
> ]     <dcterms:identifier>URN:ISBN:1850877019</dcterms:identifier>
> ]
> ] Both are correct usage of dcterms:identifier (apart from the
> uppercase
> ] as you pointed out). In both cases the identifier is the (literal)
> ] string of characters. The second identifier is not just "constructed
> ] like a URI", it is a URI!
> 
> Both are correct in that they conform to the range of
> dcterms:identifier. The problem of the first is that it isn't obvious
> what kind of identifier it is -- as it turns out it is the British
> National Bibliography identifier (I think).
> 
> The second is not a URI, it's a literal. It won't be indexed in stores
> as a URI, the sparql isURI and isLITERAL tests will treat it as a
> literal, etc.
> 
> The reason to have urn:isbn: is so that it is at least possible to tell
> what kind of a number that is, probably better to use a dedicated
> predicate like bibo:isbn in which you don't need to fudge it with a
> urn:isbn: prefix.
> 
> ] Yes, it may be confusing that you can use the URI in two ways: either
> as
> ] the (literal) string itself like here in dcterms:identifier, or as a
> ] pointer to a resource which you would do if you used it in
> ] dcterms:relation, for example.
> ]
> ] Both
> ]
> ]     <dcterms:identifier>someURI</dcterms:identifier>
> ]
> ] and
> ]
> ]     <dcterms:relation rdf:resource="someURI" />
> ]
> ] are correct -- the URI just plays different roles.
> 
> Correct in terms of "are valid RDF" and "conform to the defined
> semantics" but apart from some obscure cases involving reification, I
> think it is not a good idea to have literals that look like URIs
> without very good reason.
> 
> Cheers,
> -w
> --
> William Waites
> http://eris.okfn.org/ww/foaf#i
> 9C7E F636 52F6 1004 E40A  E565 98E3 BBF3 8320 7664
> 