Re: [open-bibliography] MARC Codes for Forms of Musical Composition

At Wed, 7 Jul 2010 17:23:29 -0400,
Houghton,Andrew wrote:
> 
> You are arguing that they are the same resource by having the
> individual, identified by the same URI, have multiple rdf:type's.
> There is a direct analogy with the TAG GenericResource-53 decision
> which basically says use separate URIs for the generic resource and
> its variant representations... per my prior example:
>
> […]

Hi Andrew,

This is an interesting analogy, but I am not sure it applies. [1]
First, a content-negotiation system for language/media type exists &
is widely deployed; no such beast exists for RDF. Second, RDF is
intended to mix & match; you would not argue that we should use a
different resource for a subject in English or in French, you would
skos:prefLabel with language tagged literals. Third, you add an
additional layer of references through which queries, etc., must work;
this is not so much a problem in the human-readable web, but it
becomes onerous to query. For example, as I mentioned before:

  isbn:123#bibo a bibo:Book .
  isbn:123#bibo rdfs:seeAlso isbn:123#vocab .
  foaf:me ex:wants isbn:123#bibo .
  foaf:you ex:owns isbn:123#vocab .

Now we have to deal with that extra seeAlso arc in order to understand
that you own the book that I want.

It seems to me that one should think carefully both before assigning a
URI multiple rdf:types and before assigning multiple URIs to what
seems to be the “same”. One can (almost) always distinguish between
any two things, but it doesn’t always make sense to do so.

> I think we can agree to disagree on this issue. OWL DL permits you
> to have multiple rdf:type's, whether that is a good idea or not
> remains to be seen. I'm saying that by limiting the individuals in
> your ontology to one rdf:type, you avoid conflating properties
> between these *variant representations* and it's still OWL DL
> compliant. I think you are saying that its too onerous to create
> separate URIs and you don't really care about the use cases where an
> application wants to explicitly know which properties are associated
> with what rdf:type since they have been conflated into a blob of
> properties for the individual.

I have not heard the use cases. I would like to. (That is, use cases
for the position that every different rdf:type should have its own
URI, e.g. http://purl.org/NET/book/vocab#Book vs.
http://purl.org/ontology/bibo/Book) Thanks!

best, Erik

1. For what it’s worth, I have argued strongly for using different
URIs to identify different resources with a difference content type.
See http://thread.gmane.org/gmane.comp.web.services.rest/8717/focus=8821

Received on Friday, 9 July 2010 05:19:40 UTC