Mobile applications are deployed in highly dynamic environments. Devices have limited resources available and the user context changes frequently. This introduces new requirements for applications, and requires that applications are able to adapt during runtime. In addition, developers must be able to incrementally add new behaviour to applications as required by unanticipated situations. Current approaches to mobile application development and architectures do not properly address these requirements of mobile applications. This work proposes a framework, based on coordination principles, which is able to facilitate the required methods of adaptation. The framework incorporates a methodology and an architecture. The architecture provides significant advantages over previous work in terms of adaptation support. The methodology provides developers with a development process and guidelines whereby adaptive applications may be specified. As new requirements of existing applications emerge, developers can build and deploy additions to applications during runtime. The research has been evaluated through the design of application prototypes. These were built through the use of an implemented toolkit for application development, which is based on the specified methodology. The architecture, methodology and toolkit provide a unifying framework for mobile applications. The work presented in this thesis closes a gap in existing knowledge in the design and execution of distributed mobile applications.

[pt] Com a evolução da Internet e o crescente sucesso de ambientes Web como plataformas genéricas de aplicação, expandiram-se as necessidades de integração e coordenação de componentes de software heterogêneos e autônomos. Os requisitos específicos deste contexto impulsionaram uma nova leva de trabalhos sobre mecanismos de coordenação, dentre os quais o espaço de tuplas reativo. Neste trabalho apresentamos LuaTS, um espaço de tuplas reativo com uma dinâmica de execução orientada a eventos e com suporte apenas a chamadas assíncronas. Desenvolvido em Lua, o sistema incorpora funcionalidades diferenciadas que permitem ao programador expandir a sua semântica básica e flexibilizam o processo de busca e recuperação de tuplas. Descrevemos a implementação de LuaTS e exploramos suas particularidades através de alguns exemplos, incluindo problemas clássicos de programação concorrente e distribuída. Conforme ilustramos, os benefícios do desacoplamento proporcionado pelo modelo de espaço de tuplas, aliados à dinâmica de programação orientada a eventos, simplificam a sincronização inter-processos e geram um fluxo de execução mais claro, reduzindo em muitos casos o esforço associado ao desenvolvimento de aplicações distribuídas. / [en] The widespread use of the Internet along with the rapid growth and acceptance of the Web as a general application platform impose new requirements associated with the integration and coordination of autonomous and heterogeneous software components. The specific needs of this context led to the development of new coordination mechanisms, among which the reactive tuple space. In this work we present LuaTS, a reactive, event oriented tuple space that supports only asynchronous calls. The system, developed in Lua, provides functionalities that allow programmers to extend its basic semantics and also support a more flexible tuple search and retrieval process. We describe the implementation of LuaTS and explore its features through different examples that include classic concurrent and distributed programming problems. We show that the uncoupling provided by the tuple space model, together with an event oriented programming dynamics,simplify inter- process synchronization and yield a clear execution stream, improving, in many cases, the development process of distributed applications.

[pt] A tecnologia de software está passando por uma transição de arquiteturas monolíticas e coesas para arquiteturas compostas por agentes e sistemas multiagentes semi-autônomos e heterogêneos. Estas arquiteturas são marcadas pela existência de propriedades adicionais no nível do sistema, como comunicação, coordenação, mobilidade, adaptabilidade e persistência. No entanto, o gerenciamento e controle destas propriedades e dos diferentes agentes de software não é uma tarefa simples e exige uma abordagem mais sofisticada desde a fase arquitetural deste tipo de sistema. Neste contexto, neste trabalho, propomos a definição de um padrão arquitetural, denominado Reflective Blackboard. Este padrão é o resultado da composição de dois outros padrões arquiteturais bem conhecidos: o padrão Blackboard e o padrão Reflection. Através da utilização do padrão, uma melhor separação de responsabilidades é atingida pelo tratamento separado, por meio de técnicas de reflexão computacional, das diferentes estratégias de controle incorporadas no sistema. De forma complementar à definição do padrão foi desenvolvida uma infra-estrutura para o desenvolvimento de sistemas multi-agentes que utilizam a arquitetura proposta pelo padrão apresentado. Esta infra- estrutura, além de funcionar como uma implementação do padrão Reflective Blackboard, provê suporte a diferentes propriedades do nível do sistema como mobilidade, comunicação, coordenação e persistência. A utilização do padrão e da infraestrutura desenvolvida foi aplicada em uma aplicação de Marketplace com o objetivo de validar os conceitos propostos. / [en] Software technology is undergoing a transition from monolithic architectures, constructed according to a single overall design, into open architectures composed of conglomerates of collaborative, heterogeneous, and independently designed agents and multi-agent systems (MAS). These architectures are driven by additional system- wide properties, such as communication, coordination, adaptability, mobility and persistence. However, controlling these properties and managing different application agents is not an easy task, requiring a more sophisticated software architectural approach. In this context, we propose the Reflective Blackboard architectural pattern, which is the result of the composition of tho other well-known architectural patterns: the Blackboard pattern and the Reflection pattern. The pattern allows a better separation of concerns, supporting the separate handling of control strategies, by means of the computational reflection technique. An infrastructure, which implements the proposed pattern and provides support to multiple system-level properties, such as mobility, communication, coordination and persistence, was also developed. The Reflective Blackboard pattern was applied together with this infrastructure to the development of a Marketplace application in order to validate de concepts proposed in this work.

Enforcing security policies to distributed systems is difficult, in particular, when a system contains untrusted components. We designed AspectKE*, a distributed AOP language based on a tuple space, to tackle this issue. In AspectKE*, aspects can enforce access control policies that depend on future behavior of running processes. One of the key language features is the predicates and functions that extract results of static program analysis, which are useful for defining security aspects that have to know about future behavior of a program. AspectKE* also provides a novel variable binding mechanism for pointcuts, so that pointcuts can uniformly specify join points based on both static and dynamic information about the program. Our implementation strategy performs fundamental static analysis at load-time, so as to retain runtime overheads minimal. We implemented a compiler for AspectKE*, and demonstrate usefulness of AspectKE* through a security aspect for a distributed chat system.