W3C home > Mailing lists > Public > w3c-rdfcore-wg@w3.org > August 2001

Re: Where's the S? (was: Re: RDFS WD snapshot frozen: http://www.w3.org/2001/sw/RDFCore/Schema/20010618/)

From: Dan Brickley <danbri@w3.org>
Date: Thu, 16 Aug 2001 19:11:03 -0400 (EDT)
To: pat hayes <phayes@ai.uwf.edu>
cc: <w3c-rdfcore-wg@w3.org>
Message-ID: <Pine.LNX.4.30.0108161858290.10586-100000@tux.w3.org>
On Thu, 16 Aug 2001, pat hayes wrote:

> Dan Brickley <danbri@w3.org>:
> >Hi
> >
> >I've frozen a snapshot of the RDF Schema spec WD that I'm working on:
> >
> >	http://www.w3.org/2001/sw/RDFCore/Schema/20010618/
> >
> >I'll use a new URL for any future versions of the text.
>
> OK, this (and Dan C's comments on the model theory) prompts a larger
> question I have for everyone: how separate/separable are RDF and
> RDFS? Should we regard them as essentially one integrated language,
> or as two languages, one an extension of the other?  Or is RDFS just
> one particular RDF vocabulary?
>
> More particularly, why are rdf:Property and rdf:type defined in the
> RDFS documentation? And why do Bag, Seq and Alt have an rdf: prefix
> but Container has an rdfs: prefix?  Why is Literal prefixed with
> rdfs: but Statement with rdf: ?

This is where the historians come in... ;-)

The answer is partly historical accident, partly layered design. In 1997
we had two RDF groups. The first, Model and Syntax, gave us the basic
triples / graph model, plus a syntax that used the XML and XML Namespaces
specs (which were themselves still works in progress, taking input from
RDF and other efforts). The second RDF WG began slightly later; the Schema
WGs job was to build on the bare bones provided in M+S, providing
additional machinery to support the specification/description/definition
of vocabularies for use in RDF applications. RDFS, in turn, tried to be
pretty minimalist, leaving fancier stuff to future work (aka "WebOnt" in
recent terminology). The groups did their best to partition the work, but
didn't really stand back collectively at any point any look at the
complete RDFS + M+S design and consider a repartitioning.

The M+S spec admits that resources fall into types (classes), since it
needed this when describing various of the features in the core spec. It
admits that properties are resources, named with URIs and hence
describable in RDF. But it doesn't go into much detail about relations
between classes and other classes; or properties and other properties. Or
constraints relating classes and properties. RDFS does that, and not a lot
else.

Some of the earlier RDFS designs were a bit more complex than the current
spec. Given the bare bones design we ended up with, some of us have
wondered whether it might be folded into "the model spec". It is indeed
odd to have rdf:type but rdfs:Class; rdf:Property but rdfs:domain. Uglier
still, IMHO, is the M+S practice of having purely XML-syntactic constructs
(rdf:resource, rdf:about) in the same namespace as the basic graph model
concepts.

My inclination for the next batch of Working Drafts is to punt on the
problem, and stick with current practice, flagging this up as something
the WG are considering.

Dan

(from web terminal in boston airport, again)
Received on Thursday, 16 August 2001 19:11:03 EDT

This archive was generated by hypermail pre-2.1.9 : Wednesday, 3 September 2003 09:38:46 EDT