Re: deciding on rdf:PlainLiteral this week

Summary:

This all may or may not matter.    If it does matter then the document
wording will have to be adjusted.  If it does not matter then adjusting
the dcoument wording should be innocuous.  So, let Sandro know your
desired wording and have him make the edits.  (I'm in an all day meeting
today.)




From: "Seaborne, Andy" <andy.seaborne@hp.com>
Subject: RE: deciding on rdf:PlainLiteral this week 
Date: Mon, 1 Jun 2009 12:05:43 -0500

[...]

> Data:
> 
> :x :p "foo" .
> 
> Case 1:
> 
> Read into RDF simple entailment; binding ?v to "foo"@en and DATATYPE(?v)
> is xsd:string.

The SPARQL function may return xsd:string on this, but this is not
sanctioned by the RDF semantics.

> Case 2:
> 
> Read into OWL2 processor, which applies the translation internally to
> "foo@"^^rdf:PlainLiteral.  

Why would an OWL 2 processor do this (except if it first translated into
the structural specification and then back out, in any case the OWL 2
specs will be changed to require the use of plain literals in RDF graph
syntaxes in this case)?

> Now query with SPARQL Basic Graph Matching.
> This defines a framework, amongst other things, for new datatypes.  But
> this datatype spans plain literals as well in a way that was not
> described before.
> 
> Inside the SPARQL engine, does DATATYPE(?v) now give "rdf:PlainLiteral"?
> At one time, the answer from an editor was "yes".

It may have been, but the whole idea of the changes to the
rdf:PlainLiteral documents was to make this not be so.

> Is that still the case?  If so, there is a change and so results of a
> query can change visibly (e.g. number of results) from the outside of
> the SPARQL processor.

Well, there are lots of similar cases that might happen in various
processors.  What make this particular case so special?

> Example: 
> ... FILTER (DATATYPE(?v) = xsd:string) ...
> Or 
> ... FILTER (DATATYPE(?v) = rdf:PlainLiteral) ...
> 
> ----
> 
> Text that does not cover it in my understanding:
> Sec 1:
> """
> rdf:PlainLiteral literals are written as RDF plain literals in RDF and
> SPARQL syntaxes.
> """
> Not about SPARQL syntax.

Well somehow the OWL 2 tool is going to have to interact with the SPARQL
tool.  If this is done through RDF syntaxes then the 


> Sec 4:
> """
> the form of rdf:PlainLiteral literals in syntaxes for RDF graphs and for
> SPARQL is the already existing syntax for the corresponding plain
> literal, not the syntax for a typed literal.
> """
> Not about SPARQL syntax.
> 
> Sec 4:
> """
> this datatype MUST use plain literals (instead of rdf:PlainLiteral typed
> literals) whenever a syntax for plain literals is provided, such as in
> existing syntaxes for RDF graphs and SPARQL results.
> """
> Not about results but close - no "syntax for a plain literal" is being
> provided - that would happen in the "SPARQL results" which are RDF
> graphs or SPARQL Query Results XML Format.
> 
> 
> I suggested explicit mention of SPARQL extended BGP matching.  Either
> that is necessary or if it is already covered, and does not no harm so I
> don't see the problem of including it.  But I'm met with strong pushback
> on that.

I think that the "pushback" is trying to understand why they "syntaxes"
doesn't cut it, not necessarily that it shouldn't be forbidden.

> Please quote text from the rdf:PlainLiteral draft that is supposed to
> cover this.  There was a proposal using "binding" but I don’t see this
> in the draft.  It may be covered in some other way.  Or maybe it's
> impossible to define a SPARQL extended matching that is preserves
> compatibility of results across an RDF view and an OWL view.
>  
> Now I may well have missed the text in the doc, or a subsequent
> discussion in the many emails.  Just point me at the text - isn't that a
> reasonable request as a response to a comment?

The problem is that I thought that the current wording was sufficient to
catch all cases, and is considerably stronger than previous wording.

>  Andy
> 

peter

Received on Monday, 1 June 2009 17:43:08 UTC