Researches On Software Evolution And Dynamism Under The Circumstances Of Internet

In this dissertation, the domain of software evolution and dynamism has been investigated and studied. It mainly focuses on several problems produced by introducing new computing model-agent (including software evolution based on agent model, modeling and analyzing multi-agents system and dynamism of multi-agents system), and software evolution caused by change of computing environment and introduction of new requirement is also discussed.Firstly, an architecture description language suitable for multi-agents system, MA ADL, has been proposed, which based on CSP formalism. In MA ADL, the organization and behaviors of multi-agents system are specified and analyzed on four different abstraction levels: component, agent, organization and system.Secondly, from a perspective of software architecture, the dynamism of multi-agents system is defined on four different dimensions, such as 'agent', 'role', 'location of agent on network' and 'communication channel between agent and the role it plays'. Through decomposing the system onto four different dimensions, the span of multi-agents system can be divided into static segment and dynamic change segment. By using this methodology, the dynamic change model of multi-agents system supported by MA ADL is established, and the algorithms of dynamic reconfiguration are also developed. Thirdly, a new dynamic reconfiguration method based on agent model is proposed. In this method, the transaction dependency relationships among agents are defined on the level of roles, which agents play. With the efforts in the phase of agent-oriented modeling and analyzing, mutual dependency and cyclic dependency relationship can be reduced to linear dependency ones. So, as for system using dependent transaction, the dynamic adaptation process of target node's reaching quiescence is determined by the linear dependency relationship of role specific agents.Lastly, the problem of software evolution caused by introducing agent model is studied based on an architecture description language named C2 ADL and software architecture type theory. In addition, the principle and implication of 'software evolution caused by change of computing environment' is also discussed. Several cases show that the legacy system can adapt to new environment through introducing new requirement and adapting its low-level technology.