W3C home > Mailing lists > Public > w3c-rdfcore-wg@w3.org > October 2002

defining RDF graph syntax

From: pat hayes <phayes@ai.uwf.edu>
Date: Mon, 28 Oct 2002 17:32:05 -0600
Message-Id: <p05111b15b9e3736c1e35@[]>
To: w3c-rdfcore-wg@w3.org

Guys, I know I said I would alter the MT to conform to the other 
docs, but after reading the other docs and the incoming comments, I 
really think that it might be worth standardizing on the simpler view 
of the RDF graph syntax that we already had worked out. Let me 
suggest that we all say some verbal variation on the following.

1. An RDF graph is a set of triples. That is the basic definition.

2. a triple consists of three parts A B C where B is a uriref, A is a 
uriref or a blank node, and C is a literal-thingie, uriref or blank 

3. We call them 'graphs' because they have a natural graphical (ie 
pictorial) rendering as a collection of nodes and arrows with labels 
written on them. But we are careful not to say that they ARE graphs 
in any mathematical sense, because they aren't.

4.. As far as the syntax is concerned, blank nodes are just anonymous 
syntactic 'things'. They are rather like existentially quantified 
variables in logic, but one can also just think of them as blobs that 
are used to connect the graph together.

5. Linear notations like Ntriples and XML use bnodeIDs to keep track 
of the blobs, but those ID s shouldn't be thought of as names or 
references, and they are local to a particular document.

The basic point of this is that it does NOT distinguish between nodes 
and their labels, and this is a real advantage, I suggest, in keeping 
the exposition clear. It certainly avoids what is otherwise going to 
be a minefield of getting the exact mathematical sense of 'graph' 
correct, and since we don't need to go into this minefield, I suggest 
that we keep out of it. In an earlier draft of the MT doc I had an 
appendix which really did use graph-theory notions to explain the 
syntax, and it was unreadable.

It is fine to be informal and talk about 'labels' in the primer (or 
anywhere else) as long as one is careful to say somewhere that since 
two nodes never have the same label, that we can (and do) *formally* 
equate labelled nodes with their labels.


PS. I could try to draft rewordings/patches if other editors felt it 
might be useful, but I don't want to tread onto sacred ground or 
anything. I think it can all be done with a few sentences here and 
there, unless anyone really *wants* to get persnickety about 
distinguishing references to a node from references to the label of 
that node.

IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola               			(850)202 4440   fax
FL 32501            				(850)291 0667    cell
phayes@ai.uwf.edu	          http://www.coginst.uwf.edu/~phayes
Received on Monday, 28 October 2002 18:32:18 EST

This archive was generated by hypermail pre-2.1.9 : Wednesday, 3 September 2003 09:52:32 EDT