W3C home > Mailing lists > Public > public-swbp-wg@w3.org > October 2004

RE: ACTION review of ODL draft

From: Gary Ng <Gary.Ng@networkinference.com>
Date: Thu, 28 Oct 2004 12:02:51 -0700
Message-ID: <CFE388CECDDB1E43AB1F60136BEB49733C6B67@rome.ad.networkinference.com>
To: "Guus Schreiber" <schreiber@cs.vu.nl>, "SWBPD list" <public-swbp-wg@w3.org>

I agree with Guus, I have no request for change, and the goals, coverage
and requirements are well articulated and tracked. 

I have gone through Ch 1, 4, 5, 6 (6 is the one specifying a Description
Logic based model). Ch 2 only briefly (it is an inclusion of previous
drafts). Ch3 Design Rationale is not included in the draft.

After today's discussion with Dan Chang, much of my previous
comments/questions have been touched on, albeit some are still remained
as questions. Included below is my edited summary to date. They are
complementary to Guus' [1] and overlaps the minutes of the telecon + IRC
log [2]. In many cases, the summary puts some context/structure from
which to interpret the minutes/chat log [2]. 

Hope this is useful. 


------------------------------------------------------------------------
----
My main questions remaining:

ODM metamodel: Inter-transformation, how language feature mismatches are

               dealt with.

UML Profiles: The extent/complexity/workflow/tool-chain of their 
		applications; legacy vs new; and also downstream
apps/code
            generation.


Revised ODM draft
-----------------

Next revision will include the following:

- Metamodels of different languages 
- UML Profile for each
- Metamodels mappings (Specified using Query View Transformation (QVT))
     OWL Full<->TM, OWL Full<->ER, OWL Full->SCL, OWL Full<->UML
- Generated XMLSchema artifacts

Scope:
-----
- Ontology Design language: RDF, OWL, TopicMaps
- Modelling: UML, ER
- OWL Full as the core

Metamodels for RDFS, OWL, OWL extension to RDFS
	  addresses ontology modelling
Metamodel for topic map
	  addresses taxonomy modelling
Metamodel for SCL  
	  addresses the domain of logic theory
Metamodel for ER  
	  addresses community on conceptual modelling, very close to
ontology.

Rather than declaring something out of scope, the above was more of an
industry led consensus of what is in scope based on industrial traction
and prevalence.

It was mandated: must address RDFS, OWL-DL (or optionally Full)
Sandpiper is working on SCL: they talked about SCL will replace F-Logic
as the formal backing of OWL. Thus it was included.

Express from STEP was a little late to be included. However, it may be
proposed during the finalizing phase starting in Jan for inclusion in
the next evolution.

De-scoping DL:
--------------

Why drop DL? It is actually moved to non-normative appendix.

Took close look at DL as core, later discussion with Deb McGuiness, her
view OWL-DL is the std definition for DL, thus there is no other reason
for another DL metamodel. 

Alan: Qualified number restriction constraint in DL, is not included in
OWL-Full/DL. Thus OWL may not be enough for general transformation. 

OWL Full as superset of OWL DL. This only applies to syntax, semantics
wise there are major differences. This shall be addressed via
constraints in SCL, will have set of common constraints, on top of which
special constraints for OWL-DL if necessary.

Dan Change will take these opinions back to ODM team.


Metamodels transformation:
-------------------------

SCL can be mapped to OWL-Full, but being more expressive, the reverse
does not hold. Thus oneway transformation.

Dan Chang to feed back to ODM group the opinion that SCL is based on
KIF, designed as interchange format. May be more appropriate as the Core
meta model.



Profiles:
--------

Louise Hart doing ODM Profiles

UML <-> OWL mappings will be useful for legacy migration of UML models
of systems to OWL. Idea is to have UML tools to model and generate OWL,
RDF, topic maps. At the same time "encourage build on a legacy base and
then proceed in another meta-model".

There will be possible problems, say mapping ER to OWL-Full due to
qualified cardinality restrictions, which is not in OWL at all. 


Outlook:
--------

OMG will publish table of mappings in Jan, may be incomplete. 
By 11th Jan it must be submitted to OMG, thus End of Nov a draft will be
available for SWBPD, the month of Dec for feedback.

Cheers,

Gary

[1] http://lists.w3.org/Archives/Public/public-swbp-wg/2004Oct/0078.html
[2] http://www.w3.org/2004/10/28-swbp-irc#T14-12-43




> -----Original Message-----
> From: public-swbp-wg-request@w3.org
[mailto:public-swbp-wg-request@w3.org]
> On Behalf Of Guus Schreiber
> Sent: 14 October 2004 09:58
> To: SWBPD list
> Subject: ACTION review of ODL draft
> 
> 
> > 5. ODM / UML ONTOLOGY PROFILE
> >
> > See Evan's message:
> > http://lists.w3.org/Archives/Public/public-swbp-wg/2004Sep/0044.html
> >
> > Comments from SWBPD?  Any updates from Guus, BenB or Gary?
> 
> I Looked at the draft ODM spec"
> 
>    http://www.omg.org/docs/ontology/04-08-01.pdf
> 
> A big plus compared to (some of the) earlier drafts is that the
> ODM covers OWL Full and not just OWL DL. Otherwise, the ODM would be
of
> limited use for users who prefer either RDF Schema or OWL Full.
> 
> I've been reading mainly Chs. 5, 7 and 8.  Ch. 5 contains a very
> readable informal description of the matches and mismatches between
> the expressivity of UML and OWL. It shows the authors have a high
> level of understanding of OWL and RDF Schema (e.g. see the remark on
> p. 27 about the relation between "classifier" in UML and
> subclasses/properties in RDF/OWL).  I can recommend this chapter as an
> instructive read for people interested in the UML-RDF/OWL
> relationship.
> 
> Chs. 7 and 8 specify resp. the RDF Schema and OWL metamodels as a MOF
> M2 model. Properties are represented with MOF classes and not as
> associations. This makes sense, because of the first-class nature of
> properties in RDF/OWL. (Note: Ch. 5, p. 26 seems to suggest simple
> binary properties should be modeled as a MOF Association.)
> 
> All major RDF/OWL constructs (class, property, resource/individual,
> literal) are represented as M2 classes. I'm not a MOF expert, but this
> suggests to me that the graphical representation as a UML Profile will
> not be very distinctive/intuitive. On the other hand, it is hard to
> see an alternative.
> 
> Overall, I'm happy with the draft as it stands. At the moment I see no
> compelling reasons for change requests.
> 
> Guus
> 
> 
> --
> Free University Amsterdam, Computer Science
> De Boelelaan 1081a, 1081 HV Amsterdam, The Netherlands
> Tel: +31 20 444 7739/7718
> E-mail: schreiber@cs.vu.nl
> Home page: http://www.cs.vu.nl/~guus/
> 
> 
Received on Thursday, 28 October 2004 19:06:33 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:09:40 UTC