Re: Annotations use case

On February 8, Jim Hendler writes:
> 
> At 0:04 +0000 2/9/03, Ian Horrocks wrote:
> >On February 7, Jim Hendler writes:
> >>
> >>  At 7:42 -0800 2/7/03, Mike Dean wrote:
> >>  >  > and recommends use of XML comments instead - ugly, but I 
> >>could live with)
> >>  >
> >>  >This is problematic in that XML comments aren't preserved in
> >>  >the RDF graph.  It precludes round-tripping of content.
> >>  >
> >>  >	Mike
> >>
> >>  no, no, no - that's my whole point!  - things one wants in the graph
> >>  go in RDFS:comment -- i.e. you need them to round trip.  Things one
> >>  adds as "comments" in the traditional programming language sense
> >>  should be lost (i.e. when you roundtrip C code the comments are lost)
> >>  and thus should be in XML comments instead!
> >
> >The C code metaphor is not valid - the "exchange syntax" for C
> >programs is the source text, including comments, and not the compiled
> >code. It would not be acceptable if comments were stripped out by
> >tools (such as editors) that work with C source code.
> 
> yes, but out exchange syntax is the RDF/XML, and there's no reason 
> one would strip the comments out if sending that -- however, if 
> parsing that into a graph and then dumping the graph back to RDF, you 
> would lose the XML comments - as would be expected - you wouldn't 
> lose the rdf:comments because someone went to the trouble to put them 
> in the graph - so things that work with our "source code" (RDF/XML) 
> wouldn't strip the comments out, but "compiling/decompiling" C would 
> strip out commments

Surely you can't be suggesting that platform specific binary code is
the equivalent of our exchange syntax? I presume that "parsing that
into a graph and then dumping the graph back to RDF" would be an
integral part of any editing of the ontology using a tool like Protege
or OilEd. This is NOT the case when C source code is edited. In
reality, compilation is a one way process and code is rarely if ever
decompiled. If we want to publish and share code, then it is the
source code that is made available - we don't publish binaries and
expect people to decompile, edit and recompile if they want to
modify/reuse our code.

> (note too that we rejected the idea of roundtripping as a requirement 
> quite long ago [see minutes of ftf2 - Amsterdam] - the decision was 
> to send a pointer to the original document if it was important for it 
> to be seen in original form - so that would work for XML as well)

We agreed that a consequence of using RDF as an exchange syntax was
that information might get "shuffled", and that this would inevitably
mean some loss of tool interoperability. We didn't agree that it was
acceptable for useful/important information to be thrown away
completely.

> >Using XML comments raises all sorts of issues and seems to be in
> >conflict with declaring RDF to be the exchange syntax. It isn't easy
> >to see how XML comments would be handled by an editor, and how they
> >would be associated with different elements of an ontology. It is
> >certainly very unlikely that all tools would do this in a uniform way,
> >so we would loose any real tool interoperability - how acceptable
> >would it be if any time I use a different ontology editor I
> >loose/trash all the comments in my ontology (consider the C
> >programming analogy)?
> 
> 
> again, my point was to emphasize that rdf:comments should NOT be 
> stripped out.  However, comments that didn't effect the graph would 
> live in the source document

It is essential that we have some way of including information in
ontologies that has no impact on the machine processable meaning of
the ontology but that will not be stripped out when the ontology is
exchanged, edited etc. 

> >
> >Just look at some of the things that are typically done with comments
> >(e.g., in the DAML+OIL ontology library, where they are often used to
> >attach natural language rubrics to classes) and ask yourself (a) if
> >this could be done using XML comments (while maintaining tool
> >interoperability), and (b) if it is appropriate that such comments
> >affect entailment in the ontology language.
> 
> exactly, if I attach different natural language to two different URIs 
> they are NOT the same (see my "embarass" example) - that's my main 
> point.

But what if I change the natural language comment attached to some
class (a URI), does that change the (machine understandable) meaning
of the class? If you believe that the answer is yes, then I wonder why
we are going to so much trouble to represent semantics in a machine
understandable/processable form when we could just use natural
language.

> 
> So I guess we are agreeing :->

I think we agreed last week - let's not hope for too much of a good
thing.

Ian

> 
> >
> >Ian
> >
> >>
> >>  --
> >>  Professor James Hendler				  hendler@cs.umd.edu
> >>  Director, Semantic Web and Agent Technologies	  301-405-2696
> >>  Maryland Information and Network Dynamics Lab.	  301-405-6707 (Fax)
> >>  Univ of Maryland, College Park, MD 20742	  240-731-3822 (Cell)
> >>  http://www.cs.umd.edu/users/hendler
> >>
> 
> 
> -- 
> Professor James Hendler				  hendler@cs.umd.edu
> Director, Semantic Web and Agent Technologies	  301-405-2696
> Maryland Information and Network Dynamics Lab.	  301-405-6707 (Fax)
> Univ of Maryland, College Park, MD 20742	  240-731-3822 (Cell)
> http://www.cs.umd.edu/users/hendler

Received on Sunday, 9 February 2003 05:39:21 UTC