Re: LANG: A proposal for the layering problem

Dan Connolly wrote:
> > I am not
> > suggesting that this be the final solution to the problem, on the
> > _some day_ you may wish to formalize N3 (perhaps SWLL etc.) into a W3C
> > recommendation. I hope that the "meaning" of such documents will not be
> > constrained by today's RDF.
> How could it be otherwise? If there are bzillions of "today's RDF"
> consumers out there that understand the meaning one way,
> there's nothing I can put in a spec to change that.

I take this to mean that _in the future_ if there were to be bzillion's of
RDF consumers, then yes I agree that this would be confusing.

A proposal: the document (i.e. root) element of a document ought indicate
the intended 'meaning' of a document (as per TimBL).

When the root element is: <rdf:RDF> the _semantics_ are per RDF (as per the
namespace name bound to the "rdf" prefix).

When the root element is something else e.g. <ont:Ontology> the _semantics_
are per whatever _that specification_ intends.

Going one step further, any special rdf:parseType="ont:collection" should be
defined by the containing document element namespace.

> > Will the 'meaning' of RDF 1
> > documents change? A simple example (insert well know namespace decls)
> >
> > <rdf:RDF>
> >     <rdf:Description ID="foo">
> >         <prop xmlns="">bar</prop>
> >     </rdf:Description>
> > </rdf:RDF>
> >
> > will this document 'mean' the same under the current RDF recommendation,
> > with the hopefully soon to be released revision?
> I don't think I understand the question.

Note the rdf:Description with an unqualified "ID" attribute. This idiom was
used in many examples in the RDF-99 rec, but the latest syntax requires
qualified attributes, e.g. rdf:ID. So the triple generated under RDF-99 is
not generated under RDF-02, hence the RDF meaning changes.

With RDF's notion of meaning being the conjunction it its triples, it is
impossible to change anything in the syntax without changing the meaning of
a document. Why isn't _that_ a problem? You might say that RDF-99 had _no_
meaning, of course a document that has no meaning in 1999 which suddenly
becomes meaningful in 2002 has if nothing else, a changed meaning.

> > Not to be difficult, but we need to be reasonable. It has not bothered
> > that the 'meaning' of an RDF vs. an OWL document might be different: The
> > true meaning of an OWL document should only be known to an OWL
> Ah.. I misspoke: I meant that it would be unacceptable if
> different recommendations gave *conflicting*, i.e. *inconsistent*
> meanings to the same document. It's all very well if
> one of them just tells you more about the document.

This is a strong argument that RDF ought have 'dark triples'  which
different specs can enlighten. Of course one can always hide meaningful
structures in rdf:parseType="Literal" xml fragments, not that I would
normally advocate such a practice, but if backed into a wall...


Received on Tuesday, 26 March 2002 15:25:59 UTC