Re: RDF Concepts and Data Model document

Tim Bray wrote:

> Tim Berners-Lee wrote:
> > I think the RDF core group has to address the  mixing of  RDF  with
> > languages. Every language spec should address namespace mixing.  We as a
> > community have been rather lax about it, partly from lack of experience
> > how we want to do it.
> Hmm, this is a tricky one, I've been kicking it around my head and it's
> not trivial at all.  Do you want to just signal that *all* RDF in an
> instance is to be believed, or do you want to individually bless chunks
> of RDF that may be found here or there?  For RDDL, you'd probably like
> to say globally "Please believe any & all embedded RDF", but one can
> imagine scenarios where you'd like to bless a particular package of
> assertions.
> Also, do you want to require that such blessing be attached to the
> instance's root element?  Also not obvious.
> Anyhow, a strawman, suppose you define some new markup in the RDF
> rdf:blessAll="true" <== believe all RDF contained herein
> rdf:bless="foo"     <== believe the embedded <rdf:rdf> element
>                          with ID "foo"
> rdf:blessings="bar" <== "bar" is the ID of an <rdf:blessings>
>                          element which has an arbitrary number of
>                          <rdf:bless id="foo"/> children indicating
>                          other <rdf:rdf> elements to believe

The issue of 'embedding' RDF , which may or may not be asserted, into other
documents has been considered by the WebOnt WG at length. WebOnt has
considered the need to embed unasserted RDF within an RDF document itself!

note that this mechanism may not be required for OWL however extension
languages that use full first order logic (i.e. NOT) will need to be able to
say that something ISN'T TRUE.

Several syntactic and semantic mechanisms of indicating which triples are to
be asserted vs. unasserted have been discussed in the aforereferenced
article, as well as in the ensuing discussion.

These issues are ones that should definately be addressed by RDF itself, and
if not in this round of RECs, then ASAP ... i.e. we really need to start
work on RDF 2 to deal with exactly these issues -- this is part of the
general issue of "colors" as alluded to by . Your
proposal is very similar to this, albeit the 'belief' status is labelled
with an ID value, whereas I've suggested that it be labelled (colored) with
a full URI(ref). In that way be can assert that we believe the contents of
an RDF document/set of triples located at a given URI(ref).


Received on Sunday, 22 September 2002 10:51:03 UTC