Re: SPARQL-WG comment on rdf:text

Sandro,

Thanks for putting this into perspective and isolating the problem.

Sorry if this is naive, but I wonder, if the goal is to prevent these
literals leaking out, whether we could simply say that they
should never be written? This could be accomplished very simply,
by making the lexical space of the datatype be empty. Then there
wouldn't be any to write.

Given that the relevant values already exist as interpretations of
plain literals in any flavor of RDF interpretation, and could easily
be made to live in the value space of rdf:text, is there any
reason to ever use a literal "fou@fr"^^rdf:text to denote such a value?

If we permit these values to be written in this alternate
way, then we'll have to practice eternal vigilance just to make sure
that they don't get written in this way (making Herbrand interpretations
impossible and so on). This would be a burden or vulnerability of every
future spec.  Better to just not let the cat out of the bag.
As a class and even as a datatype, rdf:text works perfectly well, without
having its own private syntax.

Jonathan

On Fri, May 15, 2009 at 3:48 PM, Sandro Hawke <sandro@w3.org> wrote:
>
> We got a comment from the SPARQL WG on rdf:text [1].  I gather there
> were various off-list e-mails by people trying to figure out what to do
> about it.  I was CC'd on some, but didn't read them.  Let's try to keep
> the discussion on-list, and hopefully we can keep it brief.
>
> For myself, I read the SPARQL comment to be pointing out ways that
> rdf:text literals could leak out and be visible to SPARQL users.  Before
> we went to Last Call, I believe we came to an understanding (mostly in
> discussion between me and Dave Reynolds), that rdf:text literals would
> not leak out to be visible to SPARQL users, RDF users, or anyone else
> who wasn't expecting them.  Dave presented this is a requirement for HP
> to supporting the move to Last Call, and I argued for a while before
> agreeing.
>
> We understood there would need to be more wordsmithing, to make it clear
> exactly how to prevent this leaking.  The SPARQL-WG comment nicely
> pointed out some ways the leaking could occur; this is good input to the
> wordsmithing.
>
> Does anyone see a problem here that can't just be solved by this kind of
> wordsmithing about how rdf:text doesn't leak out?
>
> (I caught wind of some questions about the value and lexical spaces of
> xs:string and rdf:text, but I don't think we need to go anywhere near
> that stuff in addressing the SPARQL comment.  If it's a problem within
> OWL or within RIF, it can be addressed on their own lists.)
>
>     -- Sandro
>
> [1] http://lists.w3.org/Archives/Public/public-owl-comments/2009May/0009
>
>

Received on Friday, 15 May 2009 22:29:03 UTC