Re: Role of N-Triples (was Re: datatyping revised draft)

Graham Klyne wrote:
> 
> At 08:28 AM 6/3/02 -0400, Frank Manola wrote:
> 
> >Independently of any usage of N-Triples in the datatyping document, it
> >might be a good idea to (once again?) clarify the usage of N-triples,
> >since I disagree with Patrick's position.  The primary RDF syntax is that
> >of a *graph* (the WG decided this, and the model theory and syntax
> >documents reflect this decision). Users *should* have to learn the graph
> >model of RDF in order to understand the normative definition of RDF.    In
> >our documents, we have used *drawings* of graphs (nodes and arcs) to
> >illustrate graphs without objection;  I believe that N-Triples are merely
> >an alternative (and often more convenient) notation for presenting
> >graphs.  I repeat:  *graphs* are normative.
> 
> So far, I pretty much agree...
> 
> >   If N-Triples can be objected to because they are a non-normative
> > notation for expressing graphs, then so can arc-and-node drawings (and so
> > could a set-theoretic notation).
> 
> ... and with this.
> 
> But my position would be that the arc-and-node drawings are _not_ a
> normative representation of a graph, any more (or less) than
> N-triples.  They are useful for purposes of explication, but no software is
> obliged to generate and/or process them.  N-triples might also have such a
> use;  I'm less convinced of that, but that doesn't matter.

Part of my point was that N-triples and arc-and-node drawings are pretty
much in the same position, so I think we're sort of agreeing.  However,
your response caused me to realize that there's another distinction to
be made here, having to do with the meaning of "normative representation
of a graph".  You're using the phrase to mean "a representation that
software is obliged to generate and/or process", and that representation
is certainly important to nail down (at the moment, RDF/XML seems to be
"it").  However, what about the concept of "the notation (or those
notations) used to describe graphs in normative specifications"?  It
certainly seems plausible to talk about a notation used to describe
graphs in a normative specification (at least, a specification that
needs to talk about graphs) as being somehow normative itself (not
normative for software processing, perhaps, but possibly in some other
sense).  N-Triples (or the variant that uses Qnames instead of full
URIs) is certainly widely used in our normative specifications for
describing graphs.  They are certainly widely used in the model theory
and syntax spec.  I can't imagine trying to describe graphs using
RDF/XML in the model theory;  in the syntax spec, it would be impossible
to use RDF/XML as a substitute for N-Triples (you'd be describing how
RDF/XML got translated to RDF/XML!). 
 
> 
> In some ways, I think the most useful message the Primer can convey is one
> that firmly establishes the idea of RDF as based on a graph -- My
> recollection from learning about RDF is that many of the other important
> issues follow quite easily once that idea is established.

I agree.  

--Frank


-- 
Frank Manola                   The MITRE Corporation
202 Burlington Road, MS A345   Bedford, MA 01730-1420
mailto:fmanola@mitre.org       voice: 781-271-8147   FAX: 781-271-8752

Received on Tuesday, 4 June 2002 13:21:00 UTC