Re: EDI = Web3.0 driving MDA?

Hi Danny,

Well said, thanks!

One of the areas that we (the ODM development team) need to thank HP 
Labs for is the contribution that Jeremy Carroll and Dave Reynolds in 
particular made to the specification. They, along with Pat Hayes, 
assisted us in ensuring that the hooks for "fitting the RDF and OWL 
metamodels to the web" were provided.  As a consequence, there is a 
discussion in the RDF metamodel section of the document that talks about 
RDF graphs that span documents, documents containing multiple graphs, 
and so forth that has proven to be very helpful to a number of folks.

The ODM is really only a first step -- it provides a consistent way for 
us to take software and models developed (or reverse engineered) in UML 
and MOF (the Meta-Object Facility)/XMI (XML Metadata Interchange, or the 
XML representation for MOF) and to use them as starting points for RDF 
vocabulary / OWL ontology development. It supports interoperability 
going the other way as well -- from RDF and OWL to UML, for example.  
The UML profiles in the specification for RDF, OWL, and Topic Maps allow 
us to use UML tools to provide a consistent graphical notation 
leveraging UML semantics for developing vocabularies, topic maps, or 
ontologies.

In the SOA ABSIG at OMG, we've been talking about extending the current 
set of metamodels to support SOA development, including but not limited 
to Semantic Web Services.  Chris Harding, who also responded to this 
email thread, spoke at an OMG meeting on the work they are doing at Open 
Group, but we've also heard from David Martin at SRI, and others active 
in SWS research -- all well received at OMG.

 From a more general web architecture perspective, there are many 
members of the OMG community who are focused on enterprise software 
development and deployment, but the definition of "enterprise" has 
certainly evolved over the last ten years. There is a strong movement 
towards content delivery, information services, information technology 
in general, all with a view on the myriad emerging deployment strategies 
for content.  A case in point: Disney joined the OMG last year, with a 
view towards standardizing content management and delivery across 
business units.  Their content delivery requirements tend towards the 
extreme -- ranging from providing metadata for every ABC news clip and 
making both the clip and the metadata available to producers with ten 
minutes of airtime, to on-demand ESPN sports, to tracking the provenance 
of individual cells from classic films, repurposing content for use in 
the "bonus features" on DVDs in many languages, and on it goes.

Anyway, I think there is a lot going on at OMG that would reflect the 
'traditional, comparatively monolithic' development, but there is also a 
lot going on that is very current, embracing both the Web and an 
evolving vision for application and content development and deployment 
in general.  I'm not unique in this view -- the same guys who I find on 
telecons for various Semantic Web working groups also show up at OMG 
meetings ... wearing similar hats but different glasses, and the number 
of folks who bridge both organizations/worlds is growing, which is 
gratifying for me and a good thing for web development, the Semantic 
Web, and software engineering IMHO.

Best,

Elisa

Danny Ayers wrote:

> On 10/07/07, Elisa F. Kendall <ekendall@sandsoft.com> wrote:
>
>>
>>  Hi Michael,
>>
>>  Actually, the work we've been spearheading within OMG is not only 
>> moving
>> towards ontology-based SOA, but is actually much broader than that.   
>> The
>> Ontology Definition Metamodel (ODM), available at
>> http://www.omg.org/cgi-bin/doc?ptc/2006-10-11, provides a
>> starting point for much more general "Semantic Web Enabled Software
>> Engineering", a phrase coined by Evan Wallace (NIST) a few years ago 
>> when we
>> first started pulling some of these ideas together for workshops at 
>> ISWC.
>
>
> Personally I very much welcome the OMG work (and the SDForum event
> sounds mighty interesting), but I would just like to chip in on a
> point that seems very important.
>
> While joining together the worlds of MDA, ontology-based modelling and
> SOA has the potential to be a big step forward on the road to the
> Semantic Web, I really hope one particular aspect will be given
> appropriate consideration. This is the "Web" in Web Ontology Language,
> the "Resource" in Resource Description Framework.
>
> Without strong recognition of the Web's protocol HTTP with its core
> concepts of resources (/URIs) and methods all we have is yet another
> set of logical models for knowledge representation. I have no doubt
> about the intrinsic benefits of ontology languages for working with
> heterogenous real-world information (compared to say straight OO,
> Codd's relational model or even Prolog...), but without fully
> acknowledging the Web side, potentially the biggest benefit may be
> overlooked - the Semantic Web itself.
>
> Coming from the direction of general Web development the notion of a
> Resource-Oriented Architecture (ROA) has recently emerged. This is
> derived from the base specs of the Web, Fielding's thesis, and like
> the W3C Technical Architecture Group's WebArch document [1] has the
> validation of practical experience in Web development to date. The
> ideas are laid out in Richardson & Ruby's book RESTful Web Services
> [2] and Google knows of many related blog posts. While this book is
> lacking material on specifically Semantic Web technologies (which is
> flabbergasting, see [3]), the ROA approach is entirely consistent with
> Semantic Web Enabled Software Engineering.
>
> For the Web developer at large, ROA offers an approach to services
> that while broadly aligned with the general principles of SOA largely
> bypasses the arcane, convoluted and often not Web-friendly
> technologies (RPC!) found in the SOAPy Web Services stack.
>
> It seems to me that without recognition there is a relatively simple
> Web-oriented approach to system development (with resources and a
> uniform interface based on URIs and the verbs of HTTP featuring
> prominently), any ontology-based MDA strategy runs the risk of
> overengineering and becoming a Byzantine stack comparable to that of
> WS-*, largely disjoint from developments on the Web. This could easily
> be counter-productive to future software development in the new
> globally-networked environment.
>
> In short, I'd suggest the Semantic Web is most productively viewed as
> an extension of the current Web rather than as traditional
> (comparatively monolithic) software with shallow Web characteristics
> added as an afterthought. This is irrespective of the sophistication
> of the object modelling techniques taken in isolation. I do hope the
> good people of the OMG bear this in mind.
>
> Cheers,
> Danny.
>
> [1] http://www.w3.org/TR/webarch/
> [2] http://www.crummy.com/writing/RESTful-Web-Services/
> [3] http://blogs.sun.com/bblfish/entry/restful_web_services_the_book
>