Re: IRI meets RDF meets HTTP redirect

> d) My initial question was precisely one area where the two views of an 
> http URI collide: after a redirect. I haven't had much joy in helpful 
> responses.

I have now had a chance to think a little more about your original issue
[1].

The rdf-charmod-uris tests (which you wrote) make it pretty clear that
in RDF there is no relationship between:

  (1) http://example.com/cat
and
  (2) http://example.com/c%61t

So it seems clear to me that if I fetch your example RDF content from
(1) or (2) I'll end up with two different RDF graphs:

  (1) <http://example.com/cat#r> rdfs:comment "A resource with multiple retrieval URIs and IRIs."@en .
and
  (2) <http://example.com/c%61t#r> rdfs:comment "A resource with multiple retrieval URIs and IRIs."@en .

Basically, I don't see how the answer to your question can be other than
(c).


On the other hand, I understand your wanting (d) and maybe (a) and (b)
to be the answer.  I think we can achieve the desired result by treating
the comparison ladder as external theories of equality that can be
merged in as desired.

That is, RDF Semantics doesn't license the inference that
   
   <http://jena.sourceforge.net/test/iri/r%C3%A9sum%C3%A9#r>
   owl:sameAs
   <http://jena.sourceforge.net/test/iri/résumé#r>

but I think the RFC 3987 does.  In particular, software which does the
IRI-to-URI translation knows about that equality and can use it.

RFC 3986 licenses case folding in the URI-schema name, right?  An RDF
processor should probably keep track of where it got that equality, and
for performance reasons it might not always use it, but I think it would
be perfectly correct to use it.  This relies on understanding that
inferences can be licensed by lots of different sources.

I've long wondered whether some HTTP redirects license equality.  I
picked "303 See Other" for the resolution to httpRange-14 largely
because it clearly keeps the referants of the old and new URIs distinct.
On the other hand, 301 says "The requested resource has been assigned a
new permanent URI and any future references to this resource SHOULD use
one of the returned URIs."  To me, recommending substitution like this
maps to RDF in the form of an equality statement.  (If we get an
equality which carries with it a suggested direction of substitution,
which I understand some FOL reasoners use, the mapping would be even
more clear.)

      -- Sandro

[1] http://lists.w3.org/Archives/Public/semantic-web/2007Apr/0127

Received on Saturday, 21 April 2007 03:01:17 UTC