TopIntroduction
A basic objective of model-driven software development is to place emphasis on the model when developing software. This is a change from the current situation, in that it shifts the role of models from contemplative to productive. The goal of model-driven engineering is to define a complete life-cycle method based on the use of various models automating a seamless process from analysis to code generation (Frankel 2003). This discipline puts all the software artefacts in the right place (e.g. business models, architectural models and design patterns) and actively uses them in order to produce and deploy applications.
Models provide solutions for different types of problems: i) description of problems and their concepts, ii) validation of descriptions and concepts represented through checking and analysis techniques, iii) model transformation and generation of code, configurations, and documentation.
Separation of concerns avoids the confusion generated by combining different types of concepts. Model-driven approaches introduce solutions for specialising models for specific concerns and for interconnecting concerns based on model transformations. This approach reduces the complexity of models through specialised modelling activities that are separated. It improves communications between stakeholders by using models to support the exchange of information. However, separation of concerns often requires specialised modelling languages to describe specific concerns, and the interoperability of specialised languages requires tools integration.
MDA proposes a set of languages and technologies (Miguel et al. 2002) to construct of modelling tools that adapt MDSD to specific platforms (e.g. EJB (Enterprise Java Beans), RTSJ (Real-Time Java Specification)) and technologies (e.g. transactions, security). Standards defining such languages are: MOF (OMG 2006), QVT (OMG 2011), MOF2Text (OMG 2008a), OCL (OMG 2010b), UML (OMG 2009a), UML profiles and RAS (Reusable Asset Specification) (OMG 2005). MDSD combines these languages to create artefact infrastructures, applicable in modelling tools, to then construct MDSD environments. However the artefact’s dependence on tool’s infrastructures makes the artefacts tool-dependent and therefore application models are also tool-dependent. The MDA philosophy to avoid platform dependency based on PIM (Platform Independent Model) and PSM (Platform Specific Models) is not reflected in the development of MDA artefacts. For example, UML modelling-tool facilities, such as profile registration and support of stereotype applications based on modelling framework tools (e.g. EMF (Eclipse Modelling Framework)), make the profiles and the models that reuse the profiles tool dependent. When exporting the models, the profiles must be exported too, and manual adaptations must be done within the model, because the profiles installed in the target tool cannot be reused. This process requires extensive experience working with models and is not feasible for complex models.
From a conceptual view-point, this chapter addresses four important challenges of MDSD methods: