Architectures for software composition

本研究将研究常见软件组合体系结构中的不良交互或副作用。组合方法是软件供应商对提供定制系统的需求的响应，并使他们的系统适应快速变化的市场需求。已经开发了各种软件组合方法，包括面向服务的体系结构、插件体系结构和面向方面的方法，以满足对灵活性和可扩展性的需求。这些组合方法的共同之处在于，假定系统组件（服务、插件或方面）是正交的，也就是说，它们不应该相互交互。然而，这通常是一个不正确的假设：一个组件的行为可能会受到其他组件的影响。系统组件直接或间接地调用或相互影响的问题被称为特征交互问题。迄今为止，解决开放、分布式系统中不良交互问题的工作还很少。因此，我们在这项研究中的目标是了解新兴软件组合方法的共性，以便能够应用通用方法来检测和解决特征交互；拥有良好的方法来建模、检测和解决组件系统中的特征交互；从具体案例中抽象出解决特征交互的通用模式；并开发在线技术（例如插件交互管理器），用于解决运行时组件之间的功能交互，这些功能交互可以与当前的组合引擎集成。

This research will study undesirable interactions or side effects in common software composition architectures. Composition approaches are the response by software vendors to the need to offer tailored systems, and to adapt their systems to rapidly changing market requirements. Various software composition approaches, including service-oriented architectures, plug-in architectures, and aspect-oriented approaches, have been developed to address the push for flexibility and extensibility. These composition approaches have in common that system components (services, plug-ins, or aspects) are assumed to be orthogonal, that is, they should not interact with one another. However, this is generally an incorrect assumption: the behavior of one component may be affected by the presence of other components. The problem of system components invoking or influencing one another directly or indirectly is known as the feature interactions problem. There is to date only a small body of work addressing the problem of undesirable interactions in open, distributed systems. Our goals in this research are, hence, to understand the commonalities of emerging software composition approaches to be able apply common approaches to detecting and resolving feature interactions; to have good methodologies to model, detect, and resolve feature interactions in component systems; to abstract generic patterns for resolving feature interactions from concrete cases; and to develop on-line techniques (eg plug-in interaction managers) for resolving feature interactions between components at runtime that can be integrated with current composition engines.