Executable Models Are Inevitable,
- Posted by: Richard Katz
- Posted on: August 28 2001 01:43 EDT
according Ivar Jacobson to reports Alexandra Weber Morales from SD 2001. And nextN month the OMG pushes that reality one more rung up the ladder with its meeting on "Semantic Actions."
Is this the right way to think about testing? Or should we start from the other end with products like JUnit and JUnitEE?
Raising UML: Jason Robbins originally started ArgoUML with the intension that it be "a joy to use." Hamburg's s Gentleware, lead by Marko Boger, has been actively involved in ArgoUML development, and now continues in that tradition. Gentleware has now published a re-worked version of ArgoUML called Poseidon for UML, Community Edition 1.0. And Gentleware intends to incorporate even more designer-centric capabilities into Poseidon, including - you guessed it - executable models.
Pushing UML to eXtreme is an interview with Gentleware's Marko Boger plus a short test drive of their new Poseidon for UML Community Edition.
-- Rich Katz
- Executable models - Yes! Executable UML models - No! by phil bradley on August 30 2001 02:16 EDT
Of course executable models are inevitable. We have had them in various limited forms for quite a while. Will they be based on UML - I doubt it!
UML is a compromised amalgamation of the 3 amigos separate modelling languages. Deficiencies are fixed by adding new extensions. Rational could learn a lot from eXtreme Programmings continual re-factoring approach. Throw out of UML things that are bad or don't work, or at least deprecate them.
UMLs approach of trying to support many modelling approaches is not the route to executable models. What is needed is a modelling approach that is tight and specific, where there is only one (or at least a small number of) right way to model something - Anyone who familiar with the Warnier-Orr methodology will know what I am talking about.
Finally, the 'two-language problem and the general-impedance problem' is not a language problem, its a tools problem. Its being solved by modelling tools that are embedded in the IDE. Unfortunately for Jacobson, not Rational tools.
BTW, the real impedance 2 language problem is between Use Cases and component/class models.