Re: A basis for convergence and closure?

On 2002-02-05 16:57, "ext Jeremy Carroll" <jjc@hplb.hpl.hp.com> wrote:

> 
> I read Patrick's proposal as:
> 
> 1:  Replace rdf:type with rdf:dType for datatyping.
> 2:  Use S-P (or TDL global) allowing 0, 1 or more rdf:dType arcs.
> 3:  Use Pat's doublet approach to fix MT (bNode denotes value not pair).

Correct. With the understanding that it is S-P but
with the whole-datatype URI denotation used by TDL, not
just the *.map component.

> This presents the following less controversial issues:
> 
> 4:    What to do with simple triples <subj> <prop> "string" ?

My comments below.

> 5:    Do we allow S-A idiom?

No ;-)

> 
> May I also suggest that:
> 
> 6:    We replace rdfs:range with rdf?:dRange for datatyping.

Only if this is absolutely required in order to
get the MT to work. OK?

> 7:    Should we rename rdf:value as rdf?:dValue?

Again, only if this is absolutely required by the MT.

We should take an anally conservative view regarding
adding any extra machinery, I think.

(OK, there, I admitted I'm anal about it ;-)


> Expanding 4: into yes/no questions is made easier by the existence of the
> syntactic transform.
> 
> Let us suppose that we already have some xslt that transforms simple triples
> into a triple pair using rdf:value. See:
> 
> http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Jan/0369.html

OK, but I think this should be part of the RDF parser functionality.
It's not just an implementational detail. I'm presuming that those
XSLT transforms are part of the parser implementation.

> Then the choices are:
> 
> 8:    Is the syntactic transform applied by default or not?

I say yes.

> 9:    Is there any standard syntactic way of indicating to (not) apply the
> transform?

parseType? per each case

processing instruction? for entire instance

> 10:   If the syntactic transform is not applied then is the untransformed
> triple just treated as in M&S with no datatyping?

I say yes.

Though this is why I think the transform should apply by
default. Folks should have to explicitly choose to do
things sans-datatyping as that, I think, is the non-standard
atypical scenario.

> 10 is a critical question because most of the heat has been about S-B versus
> the TDL global idiom (e.g. the tidiness question). If we say yes to 10, we
> say that M&S is right and both S-B and TDL are wrong.

Insofar as we abandon that idiom as a datatyping idiom, true,
though I wouldn't say either is "wrong" per say.

> A dumb triple is
> simply a dumb triple with no datatyping, and no number of other triples can
> change it.

Right. If you want datatyping, use a datatyping idiom.


> I hope we can all agree on 1,2, 3 and 10(yes).
> 
> I hope we can all live with any combinations of yeses and noes on 5, 6, 7, 8
> and 9.

Yes: 1-4, 8, 10
No: 5
Maybe: 6,7 (if absolutely needed by MT)
Open: 9

Patrick

--
               
Patrick Stickler              Phone: +358 50 483 9453
Senior Research Scientist     Fax:   +358 7180 35409
Nokia Research Center         Email: patrick.stickler@nokia.com

Received on Tuesday, 5 February 2002 10:19:26 UTC