Re: Namespace evolution.

On Mon, 16 Oct 2000, Jason Diamond wrote:

> Hi.
> 
> I'm confused. Again.
> 
> The Dublin Core site contains a working draft entitled "Guidance on
> expressing the Dublin Core within the Resource Description Framework" [1].
> In it, the suggested namespace for the 15 DC properties is
> http://purl.org/dc/elements/1.0/.

Please note that the document you mention is nothing but a working draft.
It is not endorsed by the DCMI, and it was also written before the
appearance of the RDF schema language. RDF itself has become much more
expressive since then.

...
 
> What I mean by that basically boils down to this: If properties are uniquely
> identified by the concatenation of their namespace and element/attribute
> name as specified in the RDF M&S, how are we supposed to know that
> http://purl.org/dc/elements/1.0/title ==
> http://purl.org/dc/elements/1.1/title?

This is a DCMI problem, for sure, but it is even more a RDF problem. 
Please consider the fact that we today have tools available that makes it
possible to create a RDF schema definition with a point and click user
interface. How different namespaces will have appeared five years from now
with title elements with basically the same semantics as a dc 1.1 title?

The single most important problem for the semantic web initiative is to
tackle semantic diversity.
 
> Does the DC plan on releasing version 1.2 or higher? Shouldn't a version
> agnostic namespace be used for RDF purposes in case they do? (How about
> http://purl.org/dc/elements/rdf/?

Please raise this important issue on an appropriate DC mailing list, where
the likelihood that you will be heard by responsible people is higher.

> After all, both versions contain the exact same elements.) Why hasn't
> [1] been finalized? If I quit bitching about why nothing gets done and
> did it myself, would anybody listen? Or would they just redo it later? 

That's a long story, but the most important reason is that in order to do
that job well, we needed to resolve a number of other important problems.
First and foremost, we needed workable principles for qualification and a
set of endorsed qualifiers. These components are now in place. Also the
delayed approval of the RDFS had an effect.
 
> Speaking of namespace equivalency, a recent discussion on RSS-DEV [2] brings
> up the question of how we might be able to tell when two or more properties
> are "equivalent". Could an rdfs:isEquivalentTo (or such) core property be
> added to RDF Schema before it's recommendation? Or do we expect everybody to
> agree on their URI vocabularies?

Well, I suppose that such capabilities are better implemented as add-ons
to the schema language. An equivalence is just one special case, where two
pieces of semantics are possible to exchange. You will have other cases,
where one semantic is broader than another related one. In order to cope
with semantic drift and diversity, you'll most likely have to have a
complete thesaurus like structure (broader vs narrower, use vs usefor).

Yours,


Sigfrid

Received on Monday, 16 October 2000 04:37:03 UTC