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

Re: new semantics initiative

From: Patrick Stickler <patrick.stickler@nokia.com>
Date: Sun, 16 Jun 2002 14:52:51 +0300
To: Pat Hayes <phayes@ai.uwf.edu>
CC: RDF Core <w3c-rdfcore-wg@w3.org>
Message-ID: <B93256C3.16DAA%patrick.stickler@nokia.com>

On 2002-06-14 19:41, "ext patrick hayes" <phayes@ai.uwf.edu> wrote:


> ... Are you saying that when some
> software reads some RDF, that it cannot possibly tell that
> 
> http://www.w3.org/1999/02/22-rdf-syntax-ns#type
> 
> is the symbol 'type' on the normative W3C website?

In the context of RDF, a website has nothing to do with the
interpretation of that term. If the application knows
what "http://www.w3.org/1999/02/22-rdf-syntax-ns#type"
means, then it knows how to interpret it.

The fact that
dereferencing that URIref via HTTP might result in retrieving
some content has nothing whatsoever to do with the semantics
it bears.

And the relation that it might (or might not) have to some
URIref that is a prefix of its own URIref string is also
not relevant to the semantics it bears.

All RDF employs is URIrefs, complete, opaque, unparsed.

The percieved relationship between the URIref
"http://www.w3.org/1999/02/22-rdf-syntax-ns#type" and
"http://www.w3.org/1999/02/22-rdf-syntax-ns#" is residue
from the XML serialization which has *no* semantic
significance in the RDF graph whatsoever, insofar as
the latter may have corresponded to a namespace prefix
used in the construction of the former.

Now, this is not to say that the term
"http://www.w3.org/1999/02/22-rdf-syntax-ns#type" is
not defined by an authoritative specification for
the RDF Vocabulary (not namespace) owned by the W3C. And
if it happens (by coincidence) that the RDF Vocabulary
is denoted by "http://www.w3.org/1999/02/22-rdf-syntax-ns#"
and for editorial convenience that same URIref was used
as the XML Namespace prefix for serializing the term
"http://www.w3.org/1999/02/22-rdf-syntax-ns#type" as an
element or attribute name, that doesn't result in some
formal relationship between those two URIrefs in the
RDF graph based on the structural relation inherent
in the qname. It is just a coincidence motivated by editorial
convenience. Nothing more.

This is why something line rdfs:isDefinedBy is required, if
we wish to make the relationship between term and vocabulary
explicit.

There is, however, *never* a relationship between a term and
a namespace, since namespaces are strictly a mechanism of
the XML serialization and have no functional significance in
the RDF graph.

> You sure seem to 
> be saying that. I bow to your expertise, but it seems to me that if
> this is true, then we are in much deeper doo-doo than anything to do
> with dark triples.

We are only in deep doo-doo *if* we force vocabularies to
equate to XML Namespaces. So long as we (correctly) keep these
two classes of entities distinct, and realize that there can
be an N:1 relationship between XML namespace and vocabulary,
and in fact that there can be an N:1 relationship between
qname and URIref; and so long as we don't try to force
XML qname structure and semantics onto RDF, we are just fine.

URIs are the naming model for the web. RDF takes the web as its
scope of focus.

Qnames are the naming model for XML. Namespaces and qnames are
strictly mechanisms of XML, which is simply one of many encodings
used on the Web.

RDF simply uses XML qnames to "tunnel" URIrefs between agents.
RDF does not subscribe to the full structure and semantics of
XML qnames (and in that regard is rather misbehaved) and it is
meaningless to speak of 'namespaces' in the context of RDF
graph semantics.

Not only to namespaces not exist in the RDF graph, but even
where there is the illusion of their existence, in RDF/XML
it trully *is* an illusion because the qname is not carried
forward into the graph but is "collapsed" into a URIref
that discards any notion of a distinct namespace prefix.

We simply need to stop thinking that XML Namespace equates
to RDF Vocabulary. Sometimes they seem to coincide (mostly
because it makes creating the RDF/XML easier) but they
are simply not the same thing.

>> <snip>
>> Rather, the graph syntax should have an explicit mechanism
>> that 'colors' triples as asserted or unasserted,
> 
> No, it MUST NOT have such a mechanism. Any such mechanism in the
> graph is inherently non-monotonic. This whole idea is a dead horse.

Well, I'll bow to your expertise in that matter. I still don't see
how "marking" a triple as dark based on "dark" predicates is any different
from "marking" it as dark syntactically, but I guess it must be.

Cheers,

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 Sunday, 16 June 2002 07:48:37 EDT

This archive was generated by hypermail pre-2.1.9 : Wednesday, 3 September 2003 09:49:18 EDT