Abstracting Context in Event-Based Software

With the recent immersion of advanced mobile devices in the daily lives of millions, pervasive computing is becoming a reality. Developing pervasive software systems is inherently difficult though it requires to deal with heterogeneous infrastructure besides peer-to-peer communication and device mobility. Thus, programming support for such applications must provide loose coupling between different participants, as well as loose coupling between applications and communication protocols. Event-based programming provides an appealing paradigm to decouple participants, however many event-based systems and languages have hardwired communication protocols, which limits their adaptability to heterogeneous deployment environments. In this paper we address this issue by abstracting the context in which events are created and handled, including protocols used to convey, compose, and consume these events. More precisely, we propose to extend an event-based programming model with context aspects --- conspects for short. We demonstrate the elegant usage of conspects through several examples based on their implementation in EventJava and illustrate how they allow to modularize event-based pervasive software. We also provide a thorough empirical investigation of the performance overheads and benefits of conspects, including the costs of weaving them dynamically. Through empirical evaluations, we assess the benefits of conspects for switching protocols without any changes to the base applications through three case studies with EventJava. The three studies are (1) a tornado monitoring system deployed on different architectures ranging from desktop x86 to embedded LEON3, (2) a mobile social networking suite with protocols for different scenarios, and (3) the introduction of a novel adaptive communication protocol in our mobile social network suite as well as in a novel robotic swarm application.