Re: Proposal for ISSUE-12 language-tagged literals

On 20/07/11 19:36, Richard Cyganiak wrote:
> On 16 Jul 2011, at 16:52, Andy Seaborne wrote:
>> I'd rather make DATATYPE("foo"@en) be honest and say that it
>> returns datatype rdf:LangString.
>
> You cannot do so without a hack.

You've lost me.  It puts literals in the RDF graph (old speak: abstract 
syntax) and they really do have a datatype.  It then works for RIF and 
anything else built to work with RDF.

What breaks if it is a datatype?

(It can even be a derived datatype of rdf:PlainLiteral.)

We can define it as a datatype as I suggested with a L2V mapping 
restricted to the empty set (inaccessible - that does violates the 
letter of the defn of datatype in the "non-empty lexical space" part of 
the defn but nothing relies on that), or define it with no L2V, both 
being ways of stopping syntax using ^^.  Or expand the defn of datatype 
mapping to include "foo"@en forms.

	Andy

> Assuming you want to do that, the only question is whether that hack
> goes into the RDF spec or into the SPARQL spec.
 >
> I think it should go into the SPARQL spec.
>
> Assuming that the rdf:PlainLiteral spec gets updated along the lines
> I suggested in [1], then that hack would even be nicely consistent
> with OWL and RIF.
>
> If we put the hack into RDF, then the updated rdf:PlainLiteral spec
> would have to become a hack on top of another hack, and it would make
> it harder to get buy-in for updating that spec.
>
> Best, Richard
>
> [1]
> http://lists.w3.org/Archives/Public/public-rdf-wg/2011Jul/0048.html

Received on Wednesday, 20 July 2011 18:55:29 UTC