Bounded and Eventual Adaptation for Evolvable Components

Deploying dynamically evolvable software applications is a common practice nowadays. This is typically achieved by means of mechanisms capable of adapting the system components to the modifications required by the external environment. Correctness and evolvability are closely related concerns: components might evolve along time, possibly in reaction to errors, but it is most desirable that the overall system exhibits a bounded or finite amount of error states. We propose a framework for reasoning about dynamically evolvable component systems. We introduce a basic calculus with evolvability capabilities and propose two correctness properties: bounded and eventual adaptation. While bounded adaptation ensures that at most k errors will arise in future states—including those reachable as a result of dynamic reconfigurations—, eventual adaptation ensures that the system will eventually reach a state from which no other error will arise (i.e., only finitely many errors can occur). We study the (un)decidability of these two adaptation properties in six different variants of the calculus, which represent different evolvability patterns under structural and behavioral criteria.