[BLD] RIF-RDF combination

Dear all,

I created a page about the combination of RIF and RDF, and the embedding
of RDF in RIF:
http://www.w3.org/2005/rules/wg/wiki/Core/RIF-RDF_Compatibility

This page covers the technical aspects of the combination and the
embedding; not the architectural aspects such as the reference to RDF
graphs from RIF rule sets.

Unfortunately it is a bit too late for everyone to read the page before
the telephone conference today, but I will try to give an overview of
what I've done.

In an earlier e-mail [1] I sketched an embedding of RDF in RIF. The idea
behind this embedding was that the interaction between the RIF and RDF
would be defined in terms of this embedding. After thinking a bit about
it, and consulting others, I came to the conclusion that using an
embedding for defining the interaction is probably not the best way to
go for a normative specification, for the following reasons: an
embedding redefines the semantics of a language; it can be shown that
certain properties are retained when an RDF graph is embedded, but it is
unclear whether an extension of this embedding with a set of rules
behaves as expected.

Therefore, I decided to explore the possibility of defining the
semantics of the combination in terms of combined models. Not
surprisingly, it turns out that the combination can be
(straightforwardly) embedded in an RIF rule set, for reasoning. So, any
translator which can translate RIF rules to the format A of some rule
engine, can translate an RIF-RDF combination to a set of rules in the
format A.
I explored the three normative kinds of entailment defined in
[RDF-Semantics], as well as embeddings of these three. The fact that
these embeddings exist means that this approach of combined models is in
fact equivalent to the earlier approach of embeddings, with the
difference that we now have a model-theoretic justification for the
embeddings.

During the development of the page I ran into a number of issues, both
regarding the RIF-RDF combinations and the RIF language itself.  You can
find these issues throughout the page, indented and marked with the text
"Issue:". One of the main issues is the treatment of literals and
datatypes in RIF; it is not clear at the moment (at least to me) how
ill-typed literals and unknown data types are treated in RIF.
I will extract the issues on the RIF language and send these in a
separate e-mail.

Finally, I started with the definition of a subset of RDFS, based on the
subset considered in [2], which includes the RDFS ontology vocabulary
(type, subClassOf, domain, range, etc), which does not suffer from the
complications in RDFS brought about by the use of the language
constructs in the language itself, the infinite RDF vocabulary, and the
treatment of literals. This subset can be used for the exchange of data
models, without having to deal with all the complications of RDFS.


Best, Jos


[1] http://lists.w3.org/Archives/Public/public-rif-wg/2007May/0077.html
[2] http://data.semanticweb.org/conference/eswc/2007/paper-282/html