W3C home > Mailing lists > Public > www-rdf-interest@w3.org > October 2001

Re: more on a new way of thinking about RDF and RDF Schema

From: Peter F. Patel-Schneider <pfps@research.bell-labs.com>
Date: Tue, 16 Oct 2001 06:45:46 -0400
To: tpassin@home.com
Cc: www-rdf-interest@w3.org
Message-Id: <20011016064546O.pfps@research.bell-labs.com>
From: "Thomas B. Passin" <tpassin@home.com>
Subject: Re: more on a new way of thinking about RDF and RDF Schema
Date: Tue, 16 Oct 2001 00:13:41 -0400

> [Peter F. Patel-Schneider]
> 
> 
> 
> > Edge-labelled models cannot handle things like
> >
> > <foo>
> >   7
> >   <bar>5</bar>
> > </foo>
> >
> > which are allowed in XML.
> >
> You can make a perfectly good edge-labeled graph from arbitrary XML, even
> with mixed content.  Just imagine that all continuous fragments of character
> content are wrapped in imaginary <text> elements.  Each element name maps to
> an edge label.  Child element order maps to clockwise (or counter-clockwise)
> placement of the edges.  Node are unlabeled and have no content except for
> leaves at the end of "text" branches, which contain the corresponding
> character data.  Thus all nodes are anonymous.  A document starts with an
> implied (anonymous) root node.

You are correct.  I was thinking of the RDF style of labelling, where
categories are node labels and relationships are edge labels, and
(stupidly) not thinking of making all labels edge labels.   The RDF style
is less general than a style that makes all labels the same, either node or
edge labels.  (Why?  Because the RDF style places limits on the ``order''
of labels.) 

> You may object that having all the nodes anonymous is pretty strange for
> RDF.  

Not strange to me at all.  

> I reply that for arbitrary XML, you do not know what the element names
> and so forth "mean", so you can't very well apply RDF-like assignments to
> them.

Perhaps.  But let's image that we want to give ``meaning'' to element names,
either globally, like RDF(S) can, or on a type-specific basis, like
DAML+OIL can.  Then it could be very useful to start out with a model
theory for arbitrary XML that can also be used with RDF or DAML+OIL or even
other logics.

> Now obviously you can start adding specialized rules if you want to map
> certain structures differently.  For example, you could give special
> treatment to xlink arcs and to elements in the rdf: namespace.  I'm just
> pointing out that xml does map very nicely to edge-labeled graphs even
> though it is never presented this way in books and courses.

Correct, and thanks for point out my incorrect thinking.

> The edge-labeled approach is a dual for the node-labeled (DOM) approach.
> The node-labeled approach gives you unlabeled edges, the edge-labeled
> approach gives you unlabeled nodes.  RDF gives you labels for both edges and
> nodes, but only because of a set of conventions as to how to interpret
> certain xml structures.

Yes, and, in particular, only certain XML structures.

> Cheers,
> 
> Tom P

peter
Received on Tuesday, 16 October 2001 06:46:03 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:51:52 GMT