Generic Framework for Defining Domain-Specific Models

Abstract

As a result of the widespread popularity of the Unified Modeling Language (UML) (OMG, 2003-1), many companies have invested in introducing a UML-based methodology. There are many general purpose UML-based methodologies on the market today; among the most popular are UP (Jacobson, Booch & Rumbaugh, 1999), RUP (Kruchten, 2000), Catalysis (D’Souza & Wills, 1998), Select Perspective (Allen & Frost, 1998), and KOBRA (Atkinson et al., 2001). Typically, these general purpose software system development methodologies do not immediately fulfill a company’s need. Aiming to provide methodologies that may be applied in many domains and for many purposes, these general purpose methodologies typically become extensive and are perceived as overwhelming. At the same time they typically lack support for the more exclusive needs that the companies and domains encounter. Thereby, introducing a general purpose methodology in an organization commonly implies two particular challenges that at first sight seems to be contradictory. On one hand there is a problem that the general purpose methodology provides/prescribes far too much and encounters too many situations. On the other hand the general purpose methodology does not support specific modeling concepts, mechanisms, and techniques wanted by the particular company or development group. Thus, in that respect the general purpose methodology actually covers too little. This state of affairs is why lots of consultants, researchers, and others are in the business of helping companies to introduce these methodologies, as well as customizing general purpose methodologies to be appropriate for the actual company and purpose. The customization is typically tuned based on different criteria such as domain, kind of customers, quality demands, size of the company, and size of the software development teams. A common way of customizing a general purpose methodology is by removing, adding, and/or merging prescribed tasks, phases, roles, and models/artifacts of the methodology. However, even if introduction of a general purpose methodology almost always requires a customization effort, there does not seem to be any standard and formalized way of doing it.