RE: Jena/SPARQL.org bug when processing URNs



> -----Original Message-----
> From: Toby A Inkster [mailto:tai@g5n.co.uk]
> Sent: 18 September 2008 21:40
> To: RDFa; Manu Sporny
> Cc: Seaborne, Andy
> Subject: Re: Jena/SPARQL.org bug when processing URNs
>
> Manu Sporny wrote:
>
> > Shane McCarron spotted this bug in the SPARQL.org implementation while
> > running his RDFa processor against the RDFa test suite[1].
>
> Who did? ;-)
>
> http://lists.w3.org/Archives/Public/public-rdf-in-xhtml-tf/2008Aug/

> 0021.html
>
> I'm guessing that Shane spotted the same problem as he's using the
> same URI handling library.
>
> > This seems to be a bug in the SPARQL engine - can you please verify
> > that
> > it is, or let us know why it isn't? Thanks in advance :)
>
> Ralph Swick tracked down a reference in the RDF Concepts and Abstract
> Syntax which says that for the purposes of RDF, two URIs are only
> considered equivalent if they are the same, compared character by
> character as Unicode strings.

Yes - that's what sparql.org (using Jena's IRI parsing library) is doing.

RFC 2141 using the word "lexically" is confusing - they name the same thing, but clearly they are different character strings.

For example: in http, the schema name and the host name are case-insensitive, but the rest of the URI is case sensitive.  No processor can be expected to understand all possible URI schemes.

RFC 2141 is at liberty to add that extra rules but it's URN specific not general URI.  Hence, as Ralph point out, in RDF, the minimum requirement is Unicode string comparison for general URI comparison.  An RDF processor may choose to allow comparison to take on board the rules of a URI scheme - it can be thought of as entailment regime or even as a value space with a lexical to value mapping or URIs.

        Andy

>
> --
> Toby A Inkster
> <mailto:mail@tobyinkster.co.uk>
> <http://tobyinkster.co.uk>
>

Received on Thursday, 18 September 2008 21:27:47 UTC