[RDFTM] Comments on the current draft Guideline for RDF/Topic Maps Interoperability from ODM team members

Hi All,

Our comments are fairly high-level, not at the detail that Lars has 
provided, but different from his
and thus we hope also useful.  As Lars mentions, the document is 
intended to provide a mapping
between RDF and Topic Maps representation models.  The ODM provides a 
similar mapping,
albeit between OWL and Topic Maps in the ODM.

The document we reviewed cah be found at 
http://www.ontopia.net/work/guidelines.html, and
is dated March 20, 2006.


1.  The requirements section 2 is very high level, and based on our 
experience, will be difficult to
enforce.  Through the process of learning and using the MOF QVT (Meta 
Object Facility's Query /
View / Transformation language), a recently published OMG standard for 
representing transformations
across metamodels, we have determined that there are many ways to create 
transformations from
one model to another.  While the mechanisms for doing such 
transformations are fairly rigorous and
can certainly be described thoroughly, it is likely that there will be 
distinctions that are application
specific, and that it is difficult to make some of those decisions 
without an application in mind.  An
appendix in the ODM discusses the decision we made in this regard, and 
why the mappings are
not normative in the specification.

2. The guidelines claim to be aimed at RDF/TM interoperability, but some 
OWL constructs are used:
owl:inverseOf, owl:sameAs.  We don't think this makes sense.  Either 
interoperate with RDF/S or
OWL/RDFS but not just cherry pick some useful OWL constructs.  If the 
approach is to continue by
using these OWL constructs, a disclaimer should be provided to users as 
to the limitations in the
mapping from Topic Maps to OWL or use of OWL in the mapping process.

3. We certainly understand the rationale for a guided mapping.  But what 
this involves should be
made explicit. Given an arbitrary rdf/s graph, one must first annotate 
it with the appropriate rdftm
constructs, then perform the guided mapping.  Given an arbitrary topic 
map, one must similarly first
annotate it with the appropriate rdftm constructs, then perform the 
guided mapping.

Even though the target of the mapping will have inherited rdftm 
annotations, there is no guarantee
of round-tripping unless the further development of the target is 
constrained to continue to use the
rdftm annotations.

It should be noted that this is a costly exercise. The reason for 
annotation in the first place is that
there is no convenient algorithmic way to distinguish the various cases, 
so annotation is perforce a
manual process. Maintaining the target with annotations requires the 
maintainers to have capability
in both languages, and requires many more statements, thus increasing 
cost in two dimensions.

The guidelines should therefore include default mappings for 
non-annotated constructs in all cases.

4. Is it worth making use of the http://psi.topicmaps.org/iso13250 
namespace constructs? I'm
thinking especially of /glossary/scope.

5. In 5.3 it is recommended to use the 
http://www.w3.org/2006/rdftm/rfc3066 namespace for
language tags. Is this intended to be authoritative? That is, is it an 
alias for a source maintained by
the relevant standards bodies? Otherwise it risks becoming obsolescent.

6. It may be useful to take notice of the OWL - Topic Maps mapping 
informative recommendations
in the latest ODM submission, chapter 17. Comments are invited.  That 
document, which was
published April 3, 2006, is available at:  
http://www.omg.org/cgi-bin/doc?ad/06-01-01.

Thanks and best regards,

Elisa