Re: Test case #31

At 09:13 AM 8/8/2008 +0100, Toby A Inkster wrote:

>On 8 Aug 2008, at 08:44, Hausenblas, Michael wrote:
>
>>So, the problem I see here now for us is that we don't do any
>>normalization in the RDFa TC either.

correct.  Nor should we IMHO.

>> I dunno how others feel but there
>>are basically two options: either we extend ALL the RDFa TC with
>>alternatives of normalized (like we did with XMLLiteral, e.g. in TC102
>>[4]) or we clarify that issue in the syntax.

I don't see any need to say anything about this in the spec.
We could annotate the test cases to mention that canonicalization
is neither assumed nor required.

>Could this be raised as an issue with the SPARQL query engines used  
>by the test?  Shouldn't *they* realise that <urn:isbn:0752820907> and  
><urn:ISBN:0752820907> are equivalent?

I'm not finding anything in the SPARQL language spec [1] that
requires such canonicalization.  Section 4.1.1. Syntax for IRIs [2] notes
that SPARQLs IRIref is an improper superset of RDF URI references
and does not mention canonicalization.  The only mention of IRIref
comparison [3] says they are compared as SPARQL simple literals.

  [1] http://www.w3.org/TR/rdf-sparql-query/
  [2] http://www.w3.org/TR/rdf-sparql-query/#QSynIRI
  [3] http://www.w3.org/TR/rdf-sparql-query/#modOrderBy

So I believe that SPARQL relies on the specification of
RDF Concepts and Abstract Syntax which says [4]

  "Two RDF URI references are equal if and only if they compare
  as equal, character by character, as Unicode strings."

  [4] http://www.w3.org/TR/rdf-concepts/#section-Graph-URIref

>The obvious solution would of course be to not canonicalise URIs in  
>Cognition, but that seems to be done automatically by a third-party  
>Perl module I use to resolve relative URIs, and I don't especially  
>want to abandon that module.

I believe that it is inappropriate to change the case of characters
in any part of the URI when 'absolutizing' a relative URI.  See, e.g.
"When is a URI "the same URI"?" [5].

  [5] http://www.w3.org/DesignIssues/Axioms.html#canonicalization

>And speaking of the XMLLiteral stuff, the examples with SVG would be  
>a bit easier to pass if you removed *one* of the attributes of <rect/ >.
>(Because in XML the attributes may appear in either order.)

... and probably should not be canonicalized either.

Received on Thursday, 14 August 2008 14:23:05 UTC