W3C home > Mailing lists > Public > www-rdf-logic@w3.org > September 2001

Re: RDFCore Update

From: Stefan Kokkelink <skokkeli@mathematik.uni-osnabrueck.de>
Date: Mon, 03 Sep 2001 10:24:42 +0200
Message-ID: <3B933E4A.D9748356@mathematik.uni-osnabrueck.de>
To: pat hayes <phayes@ai.uwf.edu>
CC: www-rdf-logic@w3.org
pat hayes wrote:
> ...
> >Although I completely agree on this (and on the decisions
> >made by RDFCore regarding anonymous nodes and RDF graphs )
> >it should be noted that there is no 1:1 correspondence between
> >XML serializations and RDF graphs (and hence Ntriple also).
> 
> But it is onto, ie every XML serialization maps to a unique graph,
> right? 

onto? which mapping do you mean? The mapping
XML -- parse --> RDF-Graphs is neither injective 
nor surjective. It is not injective because

parse(<Description dc:title="a"/>) 

equals

parse(<Description><dc:title>a</dc:title></Description>)

It is not surjective, because there is no XML
serialization for 
<URI1> <PROP1> _:a.
<URI2> <PROP2> _:a.

The parse function may be used to define a 
an equivalence relation ~ on XML by a~b <=> parse(a)=parse(b). 
Let XML-Ser denote the set XML/~ . Then
the mapping XML-Ser >--> RDF graphs is injective.

> (If not, please give an example.) So if the semantics assigns
> a meaning to every graph, then it also does to every XML
> serialization, which is all that it needs to do in order to provide a
> semantics for  the serialization language.

Yes.

> >I think there are two possible views on this.
> >
> >1. Accept that not all RDF graphs have a XML serialization.
> 
> That would be my own preference. Is there a problem with this?

This could be a problem for developers. For example if you use
an API like Jena for creating RDF graphs, it may be 
not possible to serialize the created graphs. 

> >2. Define an RDF graph to be a 'valid' RDF graph if there
> >   exist a XML serialization.
> 
> Well, I'd prefer a different word, call it 'XML-compliant' or some
> such. But this is a different issue, to do with the acceptability of
> RDF graphs to some processor. The model theory only has to provide an
> unambiguous interpretation for every expression; if it also provides
> one for some other expressions that is harmless.

Yes, this is no problem for the model theory. I just wanted to point
out that the decisions concerning anonymous resources have some
(possibly non-harmless) effects on the mappings between graphs,
XML, model ... Currently, we have:

XML-Ser >--> RDF Graphs >-->> N-Triple 

Regards,
Stefan
Received on Monday, 3 September 2001 04:25:24 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:52:41 GMT