David Lee +1 Even simple transforms between XSD and UML reveal numerous things than can be said in one modeling language but not the other *on both sides*. The same is true with any other tuples of model description languages you can name. Somewhere in the discussions of what *could* be communicated is lost the notion of *what should not be communicated* restrictions on the information model are as likely to offer benefits in security and privacy and resilience as well as making the programming model possible. I find myself drawn more and more to the power of transforms, including perhaps, between models as an interesting way to address the “Blind Men and the Elephant” problem presented by any and all models. As noted – we don’t need the quantum physics to have a complete model. I am more interested in can we transform the Pillar to the Rope to the Wall, creating simple integration bridges between purpose-built models. tc "If something is not worth doing, it`s not worth doing well " -- Peter Drucker
From: David Lee [mailto:dlee@calldei.com] --> Peter The question is not whether we are faithfully reproducing quantum mechanics in our data model of people, devices and phone numbers. The question is whether we have a data model that is sufficient for a reasonably large enterprise to manage data about these three entities. A proper data model encompasses all the issues involved in that question. A simple hierarchical representation of those three entities does not. -------> To bring this thread back to the beginning. The original question here (IMHO) is not " is a tree (XSD/ XML) sufficient to model everything", but "can the *components* of a data model be usefully represented as trees/XSD/XML." (quotes mine) I would argue that this is provably "Yes" by giving concrete examples from $EMPLOYEER where we have a large number of "real world" (tm) customers who have successfully modeled and efficiently implemented the processes and data involved in a variety of useful and complex "real world" (tm) objects, from financial derivatives to publishing to chemical interactions to disease and drug references to huge (literally Olympic scale) sporting events to battlefield strategic operations all using XML as the basis of the underlying model and XML technologies for (some of, or most of) the processing. So if anyone wants to argue that XML and XSD cannot be used successfully to model serious and complex entities I have some facts to dispute that. Now, that said, I am *not* claiming that XML and XSD are the "Best" way to do so or are complete in themselves. "Best" is subjective so I will avoid that. "Complete" is objective and I can assert that in most cases something more than XML needs to be added to the cake mix to come up with an actual product/solution. Say, for example, an application or web services or business logic or <gasp> humans. But that is true of any modeling technology I know of. And to +1 to Michael, the "whiteboard picture" of the whole thing rarely (ever?) shows details at the XML level or tree level. Those are at lower levels ... or "implementation details" ... So this does divide the question. Does one "think of XML/XSD" while doing high level data models ? I tend to , but unless I am talking to a bunch of other XML geeks I probably wouldn't mention it ... Any more then I would try to explain the class structure of C++ objects to the CFO ... That level of modeling is done later. At least in my experience. |