Tuesday, July 06, 2004

Models and Code

Both IBM and Microsoft seem to be adopting a strategy of physical convergence between design models and software code.

The key idea is that instead of having an entirely separate repository of models, the models are simply ways of viewing the code, while the code contains all the information necessary to (re)construct the model, and to display it in any supported notation.

There are many advantages of this convergence, since it eliminates the synchronization problems of having the model repository separate from the code base.

It also helps to bridge between the two opposing development subcultures - coders and modellers. (Coders will not actively embrace modelling, but these platforms give them modelling anyway.)

But whatever IBM and Microsoft say about standards and interoperability, most people will assume that this increases the coupling between the model platform and the code platform. We seem to be moving away from the CASE vision of platform-independent models.

Where does this leave the independent modelling tool vendors (and their customers)? They will have to find a way of demonstrating the advantages of a separate and independent modelling tool/repository in a world where models and code are converging, where the line between software development and system operations is becoming blurred, and model-based development is expanding into model-based management.

No comments: