Re: possible way forward on ISSUE-69 (1.1/Full punning) and ISSUE-72 (backwards comptability)

From: Jim Hendler <hendler@cs.rpi.edu>
Subject: Re: possible way forward on ISSUE-69 (1.1/Full punning) and ISSUE-72 (backwards comptability)
Date: Thu, 24 Jan 2008 21:27:48 -0500

> Hmm, trying to decide what I think about this - if I open an OWL 1.0  
> ontology written in the normative exchange syntax (i.e. OWL/RDF),  
  	   	      	  	    	     	    	  ^^^ RDF/XML, no?
> then how does a tool decide if it is in OWL 1.1 DL or not if the  
> functional-style syntax is normative?  

In a very similar manner to the one required for OWL 1.0.  In OWL 1.0 an
RDF/XML document was in OWL DL if it could be the result of the mapping
from an OWL DL ontology in the abstract syntax to RDF graphs.  In OWL
1.1 the situation is even simpler, as the RDF mapping document provides
an explicit mapping from RDF graphs to the functional syntax, so it is
not necessary to invert the OWL 1.0 mapping.  The introduction to
"Mapping to RDF Graphs" talks about this.

> Does that mean we must have a  
> mechanism for mapping all of the OWL 1.0 stuff into the new syntax  
> just to see if it is legal?   

Well, yes, except that there is a commitment that all valid OWL 1.0 DL
serializations in RDF triples will also be valid OWL 1.1 ontologies.

> There are a number of deployed  
> ontologies in OWL 1.0 and at least some of the current tools (like  
> Swoop) can open them and then tell the user what OWL sublanguage they  
> are in (Lite, DL or Full) and in some cases, what would be needed to  
> move into DL for use with a DL reasoner.  If we make the OWL DL 1.1  
> normative syntax be the functional-syntax, wouldn't we lose this?  

No.  Why?  

Of course there are more OWL 1.1 ontologies, because of punning, for
example, so the tools would need to be tweaked.

> (i.e. when we so called "round trip", what would be round tripping  
> to, as there'd be no normative RDF format for DL if we go with the  
> below)

The situation has not changed appreciably.  A section in OWL S&AS -
http://www.w3.org/TR/owl-semantics/mapping.html#4.2 - talks about what
it means to be an OWL 1.0 DL ontology in RDF graph form, and this is
couched in terms of the DL abstract syntax, not directly in RDF graphs.

>    It may come as a surprise to some, but I might not oppose this -  
> at this point, I'm thinking that if OWL DL necessarily lives in a  
> different syntax and is unmistakable for RDF, then OWL DL will be  
> able to go into the market that has accepted previously deployed DL  
> languages, and the deployers of RDF tools will no longer need to  
> worry about compatibility with OWL DL -- we could also change the  
> layer cake to have OWL DL sitting on XML, and only OWL Full would  
> need to sit on RDF/RDFS, which would be closer to the truth than the  
> current layer cake-- in fact, we could remove "with Semantic Web  
> basis" from the subtext of OWL DL below, and this whole thing would  
> make more sense... (yes, yes, Alan I know I'm not being positive here  
> - but frankly, I'm not sure a divorce between DL and Full wouldn't be  
> better than the very uncomfortable marriage that seems to be evolving.)

Well, I don't see how the situation has changed appreciably, so very
little evolving is happening.  

Even if there was another exchange syntax for OWL 1.1, I don't see that
a divorce is needed.  

>   -JH

peter


> On Dec 13, 2007, at 11:40 AM, Peter F. Patel-Schneider wrote:
> 
> >
> > There has been discussion on compatibility between OWL 1.1 DL and OWL
> > Full as well as backwards compatibility between OWL 1.1 DL and OWL 1.0
> > DL.
> >
> > I propose the following solution, which defines what we want for
> > compatability as well as changing DL in line with the practice of
> > existing DL reasoners.  (Changes are marked with *.)
> >
> > OWL Full: RDF extension
> > - Normative syntax is RDF graphs.
> > - Normative semantics is an extension of RDFS Semantics with
> >   extra semantic conditions on RDF(S) and OWL vocabulary.
> > - Any RDF graph is a valid "input".
> >   - Some RDF graphs are ontologies,
> >     - namely those that ... owl:Ontology.
> >   - It is expected that inputs will be ontologies.
> >
> > OWL DL: Ontology Language with Semantic Web basis
> > - Normative syntax is the functional-style syntax.
> > - Normative semantics is the DL-style semantics.
> > - Only ontologies that meet all requirements of FS syntax are allowed,
> >   - including the non-local requirements related to simple roles.
> > * Annotations are not allowed on the right-hand side of entailments.
> >   - This fixes the problem noted in ISSUE-72 at the expense of  
> > limiting
> >     what sort of questions can be asked in OWL DL.
> >
> > Relationship between the two views:
> > 1/ There is a translation T from the functional-style syntax to RDF
> >    triples.
> > 2/ If O, O' are valid LHS and RHS for OWL DL entailment
> >    then O |= O' in OWL DL implies that T(O) |= T(O') in OWL Full
> >    - This is the current relationship between OWL DL and OWL Full,  
> > where
> >      there are more entailments in OWL Full than in OWL DL.
> >      It allows for punning, defining how OWL DL and OWL Full are  
> > allowed
> >      to relate in this area.
> >
> > Desirable backward-compatibility property:
> > 1/ If O, O' are valid LHS and RHS for OWL 1.1 DL entailment
> >    and O |= O' in OWL 1.0 DL iff T(O) |= T(O') in OWL 1.0 Full
> >    then O |= O' in OWL 1.1 DL iff T(O) |= T(O') in OWL 1.1 RDF
> >    - This says that we keep exact correspondence whereever possible.
> >
> >
> > Peter F. Patel-Schneider
> > Bell Labs Research

Received on Friday, 25 January 2008 14:21:35 UTC