Re: Non- and Partial-FRBR Metadata

On 9/30/10 9:37 PM, Dan Brickley wrote:
> On Sun, Sep 19, 2010 at 8:34 PM, Antoine Isaac<aisaac@few.vu.nl>  wrote:
>> I think the complexity of the SPARQL queries would be the same. What makes
>> the SPARQL query complex (in terms of graph patterns) is the properties you
>> use in the graph patterns (thus, the number of edges)---not whether the
>> nodes are blank nodes or "fully fledged" resources.
>
> Yep. The cost of bnodes is that it makes it quite a bit harder to
> subsequently super-impose extra information on the graph. After the
> RDFCore and OWL clarifications to RDF (eg. that the same real world
> thing could have multiple URIs and graph nodes) it became much more
> common to casually assign ad-hoc URIs where formerly we'd have had
> bnodes. This idiom makes adding extra triples later much easier.
> Either way I think those decisions are up to deployers, publishers,
> annotators rather than vocab designers... we won't get much
> centralised control over when bnodes are used or not used.
>


Dan (and I guess Tom who argued for minting URIs as handles),

Indeed if there are reasonable chances that some data will exist one day for the non-empty expressions, it's worth giving it a thought!

But if the use case has really not that in mind, bnodes may just be good enough. If any other organization/data sources want to define these expressions, why not let them directly relate their resources to the work and the manifestation minted by the first data source, rather than align to a dummy resource? And let formal semantics of the vocabulary of properties identify the resources, if e.g. an axiom says that only one resource can "stand between" a work and a manifestation.

Let's be clear, I won't defend bnodes for long, but I'm reluctant to let URIs proliferate in the wild if there is very little motivation for them :-)

Cheers,

Antoine

Received on Monday, 4 October 2010 20:30:20 UTC