RE: 28 Feb. Conference Call Report (meta modelling v frameworks)

[Bob Haugen <linkage@interaccess.com>]

Keith Finkelde wrote:
>Maybe we have a different frame of reference for meta levels vs frameworks 
>within a meta level.
>I have viewed your REA as a framework within  the model [application/business] level.

Clarification:  It's not my REA, I am just a software developer who uses
the REA model.  William McCarthy, who will join this list shortly, is the
"father" of REA, and the REA Ontology is being developed by McCarthy
and Guido Geerts.  (To give credit where due...)

Difference of opinion:  REA is a semantic model and now (essentially complete)
a formal ontology.  That would constitute a metamodel in my worldview,
whereas a framework (in my worldview) would be a more like a software
foundation.  However, my reasons for proposing it to ebXML go like this:

McCarthy worked through the early stages of semantic data modeling,
where people like Peter Chen and ? McLeod? struggled with problems
of schema integration (which is roughly what needs to happen in the
ebXML BP metamodel work).  By "schema integration", I mean a 
situation where you have several independently developed schemas
that need to be integrated, and you didn't start top-down with one
neat overall schema.  

In schema integration, the lesson learned by the masters over and
over was that something like an ontology is required to classify
the different concepts in the different schemas before you can
safely integrate them.  By ontology here I mean a core set of
concepts and relationships that cover at least most of the
problem domain (in this case business-to-business ecommerce).

So the REA ontology proposes these core concepts and relationships 
(quick overview in the paper McCarthy and I collaborated on last week: 
http://homepage.interaccess.com/~linkage/REA4ebXML.htm 

It is our premise that this set of concepts and relationships is both
necessary and sufficient to classify most business objects that
have to do with economic processes (and maybe more than that -
I attended a workshop at the U of Minnesota Friday where several
people with extensive ERP experience estimated that the REA
ontology could capture 80-90% of the data in an ERP system).

The ontology is deliberately minimal: for example, few subtypes,
no attributes, etc.  It is designed to have the details of different 
business systems (or different XML schemas) "linked in" at the
proper places.  

That is the reasoning that caused me to propose REA as a way
to help the ebXML schema integration process:  it doesn't solve
all problems by any means, but it will perform a necessary
foundation service (in our opinion, anyway).

If there is agreement on the need for an ontology to guide
the schema integration, then it would seem to be desirable
to look at the REA concepts and relationships and see
what you would add or take away (or propose a different
ontology).

Thanks for the comments,
Bob Haugen