Re: Comments on structural specification (was Re: document pubication schedule)

Hi Bijan and all,

The Ontology Definition Metamodel (ODM) [1] was originally developed to 
address some of the same issues that you, Boris, and others have 
identified as the motivation behind the structural specification for OWL 
1.1, among others. We agree that this is a really valuable part of the 
overall language specification, but think that this particular document 
needs more scrutiny prior to publication from implementors in 
particular, and are willing to assist with that work, as I mentioned in 
a previous email. The end result will likely necessitate a revision to 
the ODM, which should be maintained in sync with the OWL language 
development process. We think that the OMG should remain the primary 
home for some of this work, but, just as we did with the recently 
published ISO Common Logic specification [2], we would be happy to have 
the diagrams live in both places.

Some of the areas of disconnect between the current ODM and proposed 
structural specification include a well-defined relationship with RDF, 
which Jeremy Carroll, Dave Reynolds, Pat Hayes, Chris Welty, Evan 
Wallace, and others contributed to the specification.  We also 
maintained support for OWL Full, which is important for some members of 
our user base.  There are a number of implementations of the ODM already 
available, including our Visual Ontology Modeler [3], IBM's Web Ontology 
Manager and Integrated Ontology Development Toolkit [4] (among other IBM 
projects), and several open source activities [5, 6, 7].  Thus, the 
document should be reviewed not only by us (Sandpiper), but by other 
stakeholders in the OMG community.

We are comfortable with publication of the model theoretic semantics 
document, but do not believe that either the structural specification or 
MOF-based metamodel on which it depends (whose authors are members of 
the OMG Ontology PSIG, who agree that it is merely a draft, and are 
interested in participating the work) are ready to be published with 
working draft status.

Thanks,

Elisa

[1] http://www.omg.org/docs/ptc/06-10-11.pdf
[2] 
http://www.iso.org/iso/iso_catalogue/catalogue_tc/catalogue_detail.htm?csnumber=39175
[3] http://www.sandsoft.com/products.html
[4] http://alphaworks.ibm.com/topics/semantics?open&S_TACT=105AGX01&S_CMP=LP
[5] http://www.eclipse.org/m2m/atl/usecases/ODMImplementation/
[6] 
http://www.eclipse.org/modeling/mdt/eodm/docs/articles/EODM_Documentation/
[7] http://cimtool.org/


Bijan Parsia wrote:

>
> (It would be easier if you had made these fresh threads so we could  
> collect the document discussion separate from the schedule  
> discussion. Or you could have hung this off my [impr].)
>
> Thanks for the reviews. I find them helpful. The more the merrier.
>
> I have more detailed reply, but I want to point out a general  
> perspective issue. The structural specification defines the language  
> from a certain perspective. Jeremy sometimes calls this the "tree"  
> view. This perspective has some advantages for certain classes of  
> implementor (e.g., the OWL API and the KAON2 API map fairly closely  
> into this document; even Jena has an "object oriented" view, I  
> believe). It is also useful as a bridge to related languages and maps  
> straightforwardly to an XML syntax (which is really nice for some  
> less semantic web oriented applications of OWL, for making use of the  
> XML tool chain, and for avoiding some grumbling from some XML heads).
>
> Given that the Abstract Syntax view is a pretty embedded piece of the  
> current OWL (not to mention OWL 1.1) infrastructure, I think it needs  
> to be maintained. This does not mean that the current document is  
> appropriate for any end user, but, just as with the model theory, it  
> is not intended to be.
>
> There has been some quite positive implementor feedback (e.g.,  
> Matthew Horridge, who reimplemented the OWL API in the light of the  
> structural specification and used it as a basis for Protege4). There  
> are, of course, RDF based implementations (keying off the RDF mapping  
> document); TopBraid Composer comes to mind.
>
> End user documentation is very important, but so are good tools. So,  
> I ask that when we consider the formal specifications, that we keep  
> in mind the need to support implementors, language lawyers, and  
> people trying to extend OWL or design new useful services.
>
> Boris has a discussion of some of the issues he faced as an  
> implementor in response to my [impr]:
>     <http://lists.w3.org/Archives/Public/public-owl-wg/2007Oct/0085.html>
>
> Cheers,
> Bijan.
>
>
>
>

Received on Tuesday, 23 October 2007 02:55:37 UTC