Desde o PDC que é publico que a Microsoft está a tomar uma aproximação divergente da stack OMG para MDA. E isso ficou também documentado neste artigo do mesmo Keith Short que fez a sessão no PDC. Recentemente o Steve Cook (ex.IBM) e um dos “central contributer” para o UML, publicou este paper sobre Domain-Specific Modeling and Model Driven Architecture onde se pode ler coisas como:
- MDA is misnamed: it is not an architecture at all; it is a standardized approach to model-driven development based on abstraction of platform similarities. As promoted by the OMG, it does not address the broader issues involved in using integrated models, patterns, frameworks, and tools synergistically to support software product lines. Furthermore ... the fact that the MDA is based on the use of the UML and MOF specifications restricts its usefulness even more. [pág 6]
- At Microsoft, we firmly believe that modeling is an increasingly important aspect of the software development process, and we will integrate support for modeling into forthcoming releases of Microsoft Visual Studio. We believe that it is essential to design modeling languages very carefully to suit the skills of their target users: we intend to delight our users by giving them an experience of modeling that is intuitive, agile, productive, and seamless. We are targeting our first modeling products at areas that we believe will give most immediate benefit to our customers. At the recent Microsoft Professional Developers’ conference, we announced modeling tools–we call them designers–that help the developer to design and deploy distributed service-oriented applications. [pág 5]
O “designer” SOA a que ele se refere é o mesmo que já anunciei aqui. Entretanto recomendo também estes postings do Martin Fowler, Michael Platt e do Harry Pierson de onde retirei estes quotes.
Desde que conheci o ORM que comecei a achar o UML demasiado complexo para o comum dos domain experts que normalmente faz o papel de cliente (pagador) e que devia conseguir validar o modelo. A verdade é que em determinados domínios deveria ser possível usar esquemas mais intuitivos e mais adequados ao problema que se quer modelar. Assim, o que vamos ter no Visual Studio, é uma infra-estrutura para suportar diferentes designers mas cujos metadados recolhidos vão ser usados na geração e/ou execução da aplicação. Mas não se preocupem que os diagramas mais comuns (ex. class diagram) vão estar lá de base também. Penso que não é uma questão de estar contra o UML 2.0 mas sim de não querer estar constrangido pelos modelos de extensibilidade do UML.
Recomendo este paper Domain-Specific Modeling and Model Driven Architecture (pdf 10pág.) author:
- Steve Cook is a Software Architect in the Enterprise Frameworks and Tools group at Microsoft, which he joined at the beginning of 2003. Previously he was a Distinguished Engineer at IBM, whom he represented in the UML 2.0 specification process at the OMG. He has worked in the IT industry for almost 30 years, as architect, programmer, consultant and teacher, and has focused on modeling languages and tools since the 1980s. He has published a book and many papers and articles on software-related topics.