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

Re: RDF Graph questions

From: Patrick Stickler <patrick.stickler@nokia.com>
Date: Thu, 30 May 2002 12:10:08 +0300
To: ext Graham Klyne <Graham.Klyne@mimesweeper.com>, Brian McBride <bwm@hplb.hpl.hp.com>
CC: Jeremy Carroll <jjc@hplb.hpl.hp.com>, RDF Core <w3c-rdfcore-wg@w3.org>
Message-ID: <B91BC720.15B51%patrick.stickler@nokia.com>

On 2002-05-29 22:29, "ext Graham Klyne" <Graham.Klyne@MIMEsweeper.com>
wrote:

> 
> I basically agree with Brian's comments.

I also basically agree with Brian's preference for the second
option he suggests, but for other reasons.

I.e.

>> 2) the above fragments of xml are "equal".  (A different mechanism must be
>> used to determine the namespace associated with a URI.)

Firstly, insofar as I take Brian's use of "namespace" to be
synonymous with "RDF Schema", I agree that other means than
the XML Serialization must be used to make that association,
namely rdfs:isDefinedBy.

Secondly, I will assert that the presently employed RDF mapping
from XML to graph makes namespaces in the RDF/XML irrelevant
and fully opaque to the RDF model. The graph does not preserve
namespaces utilized in the RDF/XML, and does not preserve any
qnames occurring in the RDF/XML, and therefore neither namespaces
nor qnames exist in the RDF graph at all -- hence to speak about
"namespaces" in terms of URIs in an RDF graph is meaningless.

Thus, the two XML fragments are equal, because their syntactic
differences are not significant to nor maintained in the RDF graph.

> I'll also add a further comment:  even if M&S says a property must be
> associated with a namespace, it's not clear what that actually means.

Right. It seems to be based on the incorrect presumption that a
namespace is equivalent to a schema, or document model, or ontology,
which it is not. A namespace is just punctuation, and as such, has
no inherent meaning in and of itself. It does not, by its definition
or nature, uniquely identify any schema, model, ontology, etc.

It simply "steals" the globally unique quality of a URIref to
provide a unique partition for local names. That's all.

If some namespace URI actually resolves to something, that is a
coincidence and not a quality of the namespace itself. There is
no requirement that a namespace resolve to anything, or that if
it does, that what it resolves to have any relation whatsoever
to the terms grounded in that namespace or the resources denoted
by those terms.

rdfs:isDefinedBy simply captures a URI which identifies a
resource defining its subject. If that URI happens to also
be a namespace URI, that is irrelevant.

For the time being, I think all we can do is re-state that the
object of rdfs:isDefinedBy is a resource that provides some
form of definition of the subject resource.

What that resource is, and how it defines the other resource,
should be left to additional metadata defined for the specified
object. We shouldn't make rdfs:isDefinedBy say more.
 
> I am also tending to a view that a '#'-separator should be preferred, and
> possibly even inserted automatically when forming a URIref from
> URI-without-trailing-separator + fragment (Jonathan Borden's
> suggestion?).  

Unfortunately, this does not work unless URIrefs are disallowed as
namespace identifiers, as otherwise you encroach upon the fragment
syntax specified for a given MIME type.

How do you apply the above operation to the following namespace/name
pairing?

   <boo xmlns="foo://abc.com/bar#bas"/>

The URI foo://abc.com/bar#bas#boo is invalid.

I fully agree that the present method employed by RDF for
deriving URIs from qnames discards boundaries and context,
and ideally should somehow be brought to respect XML qnames
more fully, but are we really going to change it this time
around? Perhaps we should simply document the imperfect
relation between XML qname structure and RDF derived URIs
and issue a few warnings to vocabulary designers about the
potential loss of distinction and context.

I.e., namespaces and qnames in RDF/XML are just a syntactic
necessity so that predicates can be expressed as elements or
attributes, but they don't have the same function/uniqness as
namespaces/qnames in other XML content models. That's simply
the way it is. The RDF/XML captures the graph. The graph only
uses URIs, not namespaces/qnames. So namespaces/qnames have
no real significance to the RDF model, even though we
use them.

Patrick


--
               
Patrick Stickler              Phone: +358 50 483 9453
Senior Research Scientist     Fax:   +358 7180 35409
Nokia Research Center         Email: patrick.stickler@nokia.com
Received on Thursday, 30 May 2002 05:25:08 EDT

This archive was generated by hypermail pre-2.1.9 : Wednesday, 3 September 2003 09:48:16 EDT