RE: Using ontologies to compare resources (WAS: The range of the HTTP dereference function)

Joshua Allen wrote,
> Miles Sabin wrote
> >   http://www.markbaker.ca/ qua person is Mark Baker
> > 
> >   http://www.markbaker.ca/ qua image looks-like
> >      http://www.joethomas.ca/ qua image
> > 
> > where 'qua P' is a disambiguating operator taking an ambiguous
> > URI (which, because ambiguious, doesn't identify any particular
>
> That seems most appropriate to me, as well -- this is trivial to
> represent in triples, and the "qua image" is the part that would 
> be appropriate to be catalogued in an ontology somewhere.
>
> And of course, all of this only matters if we are going to let a 
> URI be overloaded to represent multiple resources, which I am not 
> so sure is a good idea (what exactly is the benefit of all that 
> added complexity?)

I don't think there's any benefit, per se. It's simply a 
recognition of the fact that URIs _are_ often ambiguous, and that 
in at least some cases we need a mechanism for expressing the 
resolution of that ambiguity.

Over the years there've been innumerable wrangles over exactly what
unique thing a URI is supposed to refer to, particularly when
candidate things aren't electronically retrievable. Dropping the
uniqueness assumption looks like it might shift that blockage,

* We no longer have to choose which of a set of candidate resources
  is the "real" resource identified. We just declare the URI
  ambiguous and leave it at that in the absence of a disambiguator.

* We can accommodate non-retrievable resources fairly smoothly.
  Typically a URI which is declared to refer to a non-retrievable
  resource (eg. http://www.markbaker.ca referring to Mark Baker)
  can be used to retrieve an associated resource (ie. the document
  GET'able at http://www.markbaker.ca). We can declare the URI
  ambiguous between the two, and note that a GET will implicitly
  resolve the ambiguity in favour of the document rather than Mark.

* We can accomodate multiple levels of abstraction smoothly too.
  We have a long-standing issue about wether a URI identifies, say,
  a bunch of bits vs. an instantaneous snapshot of a document vs.
  an document which changes over time vs. an electronic 
  representation of The Complete Works of Shakespeare vs. The 
  Complete Works of Shakespeare (in the non-instance sense) vs. a 
  web-masters favourite work of literature vs. whatever. We can 
  declare the URI ambiguous between all of these, and allow a 
  disambiguator to take up the slack.

If URIs had been used as unambiguous identifers from the outset
then none of this would be necessary ... but that's not the way
things have worked out.

Cheers,


Miles

Received on Thursday, 28 March 2002 12:32:51 UTC