| Ubiquitous computing environments stretch the requirements of traditional infrastructures used to facilitate the development of applications. Activities are often supported by collections of applications, some of which are automatically launched with little or no human intervention. This task-driven environment challenges existing application construction and data management techniques. One of the distinguishing features of ubiquitous computing is the integration of context into the system. Context, such as time, location, and situation, allows applications to adapt to the current surroundings in order to facilitate the use of the computational environment. In this thesis, we present a file system for ubiquitous computing applications that is context-aware. Context may be associated to files and directories and is used to limit the scope of available data to what is important for the current task, aggregates related material, and triggers data type conversions, thereby simplifying the complexity of applications. Contextual information is used to dynamically assist applications in finding relevant information. Novel features of the system include how the view of data adapts to the current context, how user data is imported into the local environment, and how the system adapts to device heterogeneity. We describe several applications that we have developed within our ubiquitous computing infrastructure and show how they leverage the novel features of our file system to simplify their complexity. The system is evaluated as part of a ubiquitous computing infrastructure deployed in a prototype environment to investigate issues of performance, scalability, and usability. |
