-
Notifications
You must be signed in to change notification settings - Fork 0
k@art
K@art is a generic and extensible framework for managing dynamic component-based applications. It comprises three core parts:
- A generic and extensible meta-model for describing running system at a high-level abstraction. This generic model is platform independent but can be easily mapped to reflective component models.
- A set of meta-aspects that extends the generic metamodel with constraint checking, supervising and connections with execution platforms. The aspects are introduced using Kermeta
The proposition mentions that the rule execution should be considered through distinct points of view (dimensions). The first one concerns the Event Processing Mode, which address whether the rule may handle only one event at a time (instance-oriented semantics) or a set of events (set-oriented semantics). A second dimension covers the Execution Mode, which specifies when the condition and action parts of the rule are evaluated and executed with respect to the execution of the triggering operation. The execution modes include immediate, deferred and delayed. Finally, the Activity Mode dimension determines whether the triggering rule should be executed in the same thread of the triggering operation or not.
In terms of implementation, the paper proposes that the reactive elements are realized as hierarchical components over the fractal platform. In fact, they envisage the following reactive components:
- Event component: encapsulates application components where events of interest need to be detected.
- Condition component: encapsulates application components which should be evaluated in order to compute the condition.
- Action component: encapsulates application components on which actions should be executed.
- Rule component: encapsulates event, condition and action components.
- Policy component: coordinates the execution of the rules (rule components) that comprise the policy.
Despite the extensive theoretical discussion about rules execution, the paper does not detail the integration of this aspect on the realization part. Also, it does not address how the low-level events can be combined into high-level ones. It also does not explicitly present the language that is used to write the rules.