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

Statements about patterns of triples? was RE: RDF Terminologicus

From: Jonathan Borden <jborden@mediaone.net>
Date: Fri, 5 Jan 2001 08:47:56 -0500
To: "McBride, Brian" <bwm@hplb.hpl.hp.com>, "RDF Interest \(E-mail\)" <www-rdf-interest@w3.org>
Cc: "'Graham Klyne'" <GK@Dial.pipex.com>
Message-ID: <005f01c0771e$1b7fd2e0$0201a8c0@ne.mediaone.net>
Brian,

>
> I believe that contexts have extra semantics
> over just being a collection of statements, that they need the
> machinery of the logic layer - interpretations and that sort of
> thing to do them properly.  The RDF layer alone does not have the
> machinery  - all it can do are collections of statements.

	Right, but to be clear, the RDF layer does not attempt to do more than
represent graphs. A context might be drawn as a rectangle around a group of
nodes and arcs, but what this really means in terms of RDF is a bag node
with a bunch of arcs travelling to nodes within the context.


>
> I share your concerns about the current definitions of bags - have
> ordered properties for the members of an unordered collection just
> seems bizare to me.  So I'd be sympathetic to a future spec defining
> improved collections.  But right now BAG is what we have, so in terms
> of working with the current spec, I think BAG is right.

	I completely agree. In fact, this whole discussion has shown me that the
base RDF 1.0 M&S specification is really quite robust (e.g. being able to
support contexts) but the one thing that I really dislike is the naming
specification for bag predicates i.e. _1,..._n which:

	1) impose a numerical order on an unordered collection
	2) aren't numbers and hence don't sort as numbers

	On the other hand that's what we have. Perhaps there's a way out of this by
layering additional specifications on top of RDF M&S e.g. suppose we state:

	A collection is represented by an rdf:Bag.
	All predicates matching the pattern "_*" must start with a single "_"
followed by a stringized integer > 0.
	Membership in a context is defined by triples of the pattern:
		"rdf:_*" (expand qname for URI).
>
> Contexts will bring a much richer and powerful semantics.

	I think contexts are a fairly essential concept.

Jonathan Borden
The Open Healthcare Group
http://www.openhealth.org
Received on Friday, 5 January 2001 08:44:10 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:51:47 GMT