UML and Paralysis by Analysis

Taking a break from Government complaining and back to work.

First point: UML is no silver bullet. It is only as good as the people producing the diagrams, and the people reading the diagrams. If the diagrams are produced by someone who doesn't understand the problem, how can they specify how to solve the problem? And, if they need to be specified to the implementation level, what does that say about your developers? UML works better when used in line with Brook's original hypothesis, that software should be iteratively grown. Build the parts that you can understand, and add (and rebuild) new parts once you have the framework done.

Second point: UML is usually done to death. It's similar to designing a house down to the cut lengths of each board. The additional time and effort in designing to that level is lost when an error is made. However, the error won't be determined until elaboration or construction. So what's the point in trying to design the interaction between all the components, none of which exist yet? UML should only be created to describe the problem to be solved to the level that a developer can understand it and ask questions. Using it to describe the entire application creates a waterfall model, instead of continually refining the model as the application evolves.