aboutEachPrefix: what to do?

On Wed, 13 Sep 2000, Graham Klyne wrote:

> At 02:00 AM 9/13/00 -0700, Jason Diamond wrote:
> >P.S. What the hell am I supposed to do with aboutEachPrefix? It doesn't seem
> >possible to implement.
> Another one for the issues list?

Already got this one!

	"Something should be done about aboutEachPrefix construct"
	Raised Tue, 29 Feb 2000 by mailto:timbl@w3.org 
	Summary: Is it best to put it off to a level of logic above the basic RDF? 

I'm snowed under somewhat but hope to get a new version of the Issue list
out shortly. There's been a record breaking amount of RDF traffic,
announcments, debate etc in the last week which has (I'll
admit) taken up a fair slice of my attention.

> I think this is feature that doesn't belong in either the RDF model or 
> syntax spec.

I agree. It's close to the notion of intensionally defind classes that the
RDFS schema WG toyed with but punted on.  Eg. the class of people who I
decide can have access to (say) my online family photo page. We can't pick
these out with a simple regex over URIs as in aboutEachPrefix, but some
expression couched in terms of the RDF data model could do that job. So
the question in my mind is whether we want a general, query/logic based
way of characterising classes of RDF resource, and then do aboutEachPrefix
in terms of that. Or whether the basic URI/regex based functionality of
aboutEachPrefix, re-expressed properly, is useful enough to treat
separately without waiting for fancier logic/query facilities.

> I think that either:
> (a) the functionality reasonably exists as part of a tool for gathering RDF 
> statements, or

I'm not sure quite what you mean here. Nor why the contrast with (b). 

> (b) can be modelled as an RDF property, rather than as part of the RDF core 
> specification.

Yep. We'd need to be clear whether we were talking about the relation
util:regexLeftMatch that holds between URIs considers as strings, or 
the relation util:furtherUpInSomeURINamespace that holds between
the resources named in some URI schemes. Which (as ever) drifts us back
into the Resource<->URI debate.


Received on Wednesday, 13 September 2000 09:18:04 UTC