Re: defining quality of OOA and OOD models


Phlip,

> By how well the model participates in implementing source code that passes
> all tests, is clear and expressive, duplicates no behaviors, and has minimal
> classes, methods, & lines.

OK, I see here 4 different criteria:

1) impact on the code correctness and satisfaction of all functional
requirements (OOD?)

2) clarity and expressiveness (OOD?)

3) no cloning (OOD?)

4) leanness & brevity (OOD?)

(OOD?) means I assume you were talking exclusively about OOD models

> The question of evaluating a model in isolation from code is useless. What
> would you do if the model evaluated as wonderful but the resulting code
> sucked? Is that a useful evaluation?

I did not ask about evaluating in isolation from code - evaluation with
respect to code is probably a large subset of overall evaluation
criteria that has to be applied and also consists of many subcriteria -
some of which you mentioned above...

> In which case why is it OO (meaning leveraging objects with polymorphic
> methods behind narrow interfaces)?

Just focusing on OOA as it was thought in the old days as a method for
understanding, capturing, and communicating problem domain...

Thanks for your input,

Davor

.