| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
Bill, I think we
may be talking at cross-purposes. I think
you may be asking the question whether it makes sense to have a single ontology,
which is used as a ‘lingua franca’ for the integration of systems (I had taken
this as an assumption/premise as I thought/think this is the basis upon which
the SUO is being proposed – maybe I am wrong?) – and suggesting this will not
work. That what is needed is a translation mechanism between ontologies. If you are
right, then integration has to work on a point-to-point basis. When a new
ontology is introduced it has to be translated into each of the other
ontologies (or, if you are lucky, into one of the ontologies that is more
expressive than it and piggy-back on its translations). What I am
unsure about is whether, as there is a translation, one cannot have a nominated
ontology as a ‘lingua franca’ and if not, why not? My point
was different. It was that once it has been decided to develop an integrating
ontology (I presume the one that you claim “breaks” easily), it makes practical
sense not to develop a promiscuous ontology – but to regiment it. I was not
meaning to imply – as you might have interpreted what I said – that every
system should use the integrating ontology. There will be cases where a system
has its own ontology and for integration translates this into the integrating
ontology. The point of the integrating ontology is to keep the number of
translations needed to a minimum – one for each new ontology/system – and as
simple as possible. As an analogy - at the technology level this is the
strategy of most middleware systems. Regards Chris -----Original
Message----- Chris P: > However when one tries to build or
integrate (EAI) business operational Perhaps I misunderstand your point, but I
have a different view, which is at odds with your regimentation. My own
experience seems to show that regimentation is impractical and ineffective as a
means of integration. If my "regimented" ontology you mean a
"standard" ontology used by the integrated application as the
integration mechanism, then I would argue that these ontologies
"break" very easily and quick because unlike human languages they
can't adapt to changing conditions and requirements. Personally, I see
the promisciuty of ontologies as a *good* thing that should be encouraged -
which of course pushes the integration solution in another direction: that of
mapping between ontologies. Bill ------------------------------------------------------------------ |