Peter, your thoughts are so far-reaching that a certain problem of communication easily arises, as terms acquire a weight and meaning not immediately obvious to everyone. This may apply to both words which are key in your concept, "model" and "metadata". So it is more impressions, not yet real understanding, what I manage to collect. But here I record them, and they are organized around single words.
Word "comprehensiveness". I feel that the model you are thinking about is extremely comprehensive: ready to ingest just about any metadata about just about anything (business processes, infrastructure, ...). Almost a container of metadata.
Word "layer". As a result, the model is less a single model in the conventional sense, but a layer - a new (?) "model layer", which you implicitly claim to be as fundamental as the conventional "persistence layer" and "application layer". What distinguishes a (conventional) model from a model layer? The former is a centralized and closed structure, growing by a "pull" process - additions are deliberate and selective; the latter is decentralized and open, growing by a "push" process - whenever a certain kind of artifact is created and made accessible, it is automatically added to the layer. (Yes, of course, this way of putting it reminds of the semantic web and RDF.)
Word "coherence". John's concern about the sanity of "merging them [data models] into a single data model" reflects the view of distinct models which can only be integrated by an effort of adaptation etc. From this view point a "model layer" is as incoherent as the persistence layer - a mere collection of models. But I think the core of your vision is a new kind of coherence, enabled by a uniform navigational model, which is supplied by the graph database approach. Thanks to this coherence, anything added to the model layer is immediately addressable, explorable, usable - so-to-speak "self-integrating".
Word "transformation". As your model is essetially a model of metadata, the transition to usable artifacts (e.g. WSDLs, database definitions etc.) is the result of "search+transform" - collecting the appropriate metadata, exploring what they imply, and materialize those implications into artifacts.
What I wonder about is the role of vocabularies and structural patterns. The fact of a superb navigational model
does of course not yet guarantee usefulness - it requires an understanding of the "words" you encounter during your navigation (traversals, you would say, probably). So do vocabularies, or ontologies, play a crucial role in the construction of the Enterprise ready model? And do standard structures, which would be comparable to complex types in XSD, describing the assembly of complex units from building blocks?
Thank you for giving us these glimpses. Should I be completely off the track in my attempts to approach some understanding, perhaps you give me a warning.
Hans-Juergen
(or Hans, just as you like)
