The purpose of WP3 is to develop a model-driven, dynamic aspect composition framework for co-dependent, co-existing configurations. It will address models at runtime and dynamic variability dimensions based on experimentation.

The role of adaptation rules is to define how variants should be chosen depending on the evolution of the environment. The modelling of these rules requires that beyond the modelling of variation points, interesting properties of the platform are also modelled.

Once both variation points and rules are modelled there are two options for implementing the application. The first one is to "compile" the adaptation aspects and rules into the application code. This could be done manually or using model transformations, but it has the disadvantage of hard-coding the adaptation rules at implementation-time.

The second option, which we will implement in this work package, is to keep the models at runtime and dynamically interpret them in order to supervise and adapt the behaviour of the application. This solution has the advantage of making the adaptation process both dynamically changeable and easier to monitor. It will be implemented using meta-modelling languages such as Kermeta which allows the definition of the semantics of meta-models and the execution of corresponding models.

• OSGi as the unerlying execution platform. OSGi is a widely used platform in the industry.
  
• Drools/Java for the weaving of aspect. Aspect models are now compiled into efficient code that is directly executed at runtime.
  

WP3 has demonstrated several prototypes combining Aspect-Oriented Modeling and Model-Driven Engineering techniques for managing dynamic variability:

• Models@Runtime workshop at MODELS'08 and MODELS'09
• Forum Demo at AOSD'10