The language is too complex.We need to change that. Eighty percent of all applications can be designed with less than 20% of UML. In my company we have defined a pure subset of UML to become the(c) Ivar Jacobson, один из авторов
ЪEssential Unified Modeling Language.We also use a very different way of describing UML that is much more attractive to ordinary users. Traditional UML is designed for methodologists or tool vendors.
I would love to restructure UML as a set of domain-specific languages (DSLs). I would like to do that similar to the way were designed the Unified Process in my company.A DSL is an aspect of a modeling language (of which UML is an example). You create your modeling language as a composition of many such DSLs (aspects) in a way similar to how you compose a software system from many cross-cutting concerns.While I claim that the language was not designed for users but for methodologists and for tool vendors, I claim it was not even good for the latter groups.The semantics of UML are poorly defined. UML—in particular, UML 2.0—has included so many constructs from so many different methodological camps that it became impossible to define its semantics clearly. Like many other languages, UML became, as John Backus said about Ada, “fat and flabby.”
The focus was on concrete syntax (icons) and to some extent on static semantics, but we left the operational semantics undefined. I expected that we would get this critique since standard language design practice at that time was to use techniques like denotational semantics. We didn’t. We just wrote page after page that we knew were very hard to understand.
Граждане дико жгут по моему. Кто такие эти мифические жывотные - methodologists ?