W3C home > Mailing lists > Public > w3c-rdfcore-wg@w3.org > April 2003

RE: pfps-08 last call comment on typed literals

From: <Patrick.Stickler@nokia.com>
Date: Fri, 4 Apr 2003 13:52:40 +0300
Message-ID: <A03E60B17132A84F9B4BB5EEDE57957B01B90C70@trebe006.europe.nokia.com>
To: <jjc@hplb.hpl.hp.com>, <pfps@research.bell-labs.com>
Cc: <phayes@ai.uwf.edu>, <w3c-rdfcore-wg@w3.org>, <bwm@hplb.hpl.hp.com>



> -----Original Message-----
> From: ext Jeremy Carroll [mailto:jjc@hplb.hpl.hp.com]
> Sent: 04 April, 2003 12:31
> To: Jeremy Carroll; Peter F. Patel-Schneider
> Cc: phayes@ai.uwf.edu; w3c-rdfcore-wg@w3.org; bwm@hplb.hpl.hp.com
> Subject: RE: pfps-08 last call comment on typed literals
> 
> 
> 
> Peter Patel-Schneider wrote:
> > > So finish the job!  :-)
> 
> I think I will try to echo what I hear Peter suggest, since 
> Patrick seems to
> be (1) following this thread (2) of the opinion that there is 
> not a proposal
> that changes rdf:XMLLiteral to be a normal rather than a 
> special datatype.

Well, I wouldn't be opposed to rdfs:XMLLiteral being a normal datatype,
if that essentially means that the lang tag becomes irrelevant to
equality tests. In which case, we could still then remove lang tags
from typed literals.

I've never much liked the idea of lang tags infecting XML literals,
which I consider to be fragments disjunct from the semantics of
the RDF/XML serialization itself. But I won't go there (again)...

> So:
> We have not been contemplating changing either:
> 1) the rdf:parseType="Literal" syntax
> 2) the representation of this, as a canonical XML doc, in the 
> domain of
> discourse

Agreed.

> We have  changed the treatment so that:
>   most of the syntactic transformation of from (1) to (2) is 
> done in Syntax
> and not in rdf:XMLLiteral mapping rules.
> We could go the whole hog, and make the mapping rules of 
> rdf:XMLLiteral the
> identity.
> Leaving (2) fixed we have that the syntax document not only has to
> canonicalise the literal content, but also it converts it into an XML
> document as currently specified in concepts.
> This would change test cases: e.g.
> 
> <rdf:RDF>
>   <rdf:Description xml:lang="en">
>     <rdfs:comment rdf:parseType="Literal"
>     ><b>Bold</b></rdfs:comment>
>   </rdf:Description>
> </rdf:RDF>
> 
> would correspond to n-triples
> 
> _:a rdfs:comment
> "<rdf-wrapper 
> xml:lang=\"en\"><b>Bold</b></rdf-wrapper>"^^rdf:XMLLiteral .
> 
> or something like that (I don't swear I've canonicalized correctly).
> Whether or not we include a redundant @en in the literal is 
> immaterial.

I think that requiring parsers to do the canonicalization would be
an onerous burden and not one that would be easy to sell. But since
I'm not a parser implementor, and if you and Dave think it's no
biggie, then who am I to suggest otherwise ;-)

In any case, I really wouldn't like to see <rdf-wrapper ...> in
the actual graph. I didn't much like it as a conceptual element.
I will like it less as a real element.

> We may well then have
> rdf:XMLLiteral as a subdatatype of xsd:string
> i.e. it is completely well-behaved and may or may not have 
> synonyms and we
> don't care, and language tags never participate in l2v 
> mappings. We could
> even ditch them all together from typed literals.

Well, I still think that reverting to M&S XML Literals (and ditching
lang tags from typed literals) is the better way to go all around.

It doesn't change the parseType syntax. It doesn't change the
canonical representation of XML literals. It retains the lang
tag (unfortunately) but in that way remains true to M&S.

The definition of typed literals and D-interpretations is thus
freed from "contamination" from the legacy literal types.
 
Patrick
Received on Friday, 4 April 2003 05:52:48 EST

This archive was generated by hypermail pre-2.1.9 : Wednesday, 3 September 2003 09:56:52 EDT