Re: RDF-ISSUE-13 (RDF XMLLiterals): Review RDF XML Literals [Cleanup tasks][Turtle][JSON]

On 3/9/2011 10:03 PM, Ivan Herman wrote:
> The question is whether it is ok to put this kind of load on all processes producing RDF. Because, clearly, this is not only on RDFa, but on any other process thata dumps data into RDF.
>
> Though c14n is not very complicated per se, it is not simple in practice. People may use off the shelf XML libraries to serialize a subtree and if that library does not canonicalize (and very few do), then users may have a problem... There fewer parser writers than data producers...
>
> I would say this should be revisited if possible.
>
> Thanks
>
> Ivan
>

I think the issue is where do we expect XMLLiterals to arise:

a) from RDF/XML => 2004 soln is good
b) from general code migrating legacy data into RDF without going via 
RDF/XML => 2004 soln is not good

I agree that c14n is not easily available to typical XML programmers; 
however for RDF/XML parser writers this is not a burden.
Any other format could mandate that XMLLiterals get canonicalized on 
input. e.g. Turtle could specify that
    "<foo/>"^^rdf:XMLLiteral
actually means
    "<foo></foo>"^^rdf:XMLLiteral
(I think that is correct ...)
Alternatively triple stores could be responsible for canonicalizing.
Or we canonicalize in the mapping to value space, hence in the equality 
algorithm (design rejected in 2003 last call)

Jeremy

Received on Thursday, 10 March 2011 08:20:44 UTC