OWL2RL document updated

Thanks to Axel and Jos for review comments and to Sandro for 
suggestions. I have updated the document in line with their comments.

Note that the changes I've made haven't been mechanically verified since 
I don't currently have a complete RIF PS toolchain that I can link to 
our OWL RL test harness. I've put editor's notes into the two appendices 
about this.

Jos de Bruijn wrote:
> 1- I suggest to mention the two kinds of OWL 2 RL implementations (fixed
> rulesets and template instantiation) both in the abstract and
> introduction, to make the document easier to read.  I think references
> to the appendices in the introduction would also be appropriate, for
> those readers who who are not interested in discussions of the issues,
> but rather want to see the implementation immediately.

Good idea, done.

> 2- The end of Section 2 is a bit awkward.  Some concluding statements
> would be appropriate.

Done.

> 3- section 3: which entailment notion for RIF-RDF combinations is
> referred to here?  Is this RIF-Simple-entailment? [[SWC#def-simple-entails]]

Yes, clarified in doc.

> 4- The translation does not consider plain literals with language tags;
> they should be translated typed literals with type rdf:PlainLiteral.

Isn't that catered for automatically by use of the RIF-RDF combination 
semantics?

> 5- the editor's note in section 4.4.3 has not yet been addressed.

Now done - makes the rule sets a bit bulkier. :-)

Axel Polleres wrote:
 > 1)
 > the suggestion of rif:error as a unary predicate is a good idea, but
 > conflicts with the use of rif:error as a 0-ary prediacte in the
 > RDF-OWL-compatibility document [1]

Now changed to a 0-ary predicate for consistency with a note explaining 
that implementers may choose a more diagnostic-friendly approach if they 
wish.

 > 2)
 > I am afraid there is a problem with rule (* eq-diff2 *)... It does not
 > prevent that ?x and ?y bind to the SAME member of the list, where no
 > inconsitency should apply... the solution is to extend the memeber
 > predicate with an index. and let the rule apply to differently indexed
 > members only,

Now fixed thanks.

Follow a suggestion from Sandro I'm now using the List builtins to 
replace _member (though it in this case it makes the individual rules a 
little bulkier).

At the moment I haven't replaced the other list rules with list builtins 
because I can't see how to do that neatly yet.

Dave

Received on Saturday, 26 September 2009 17:08:45 UTC