Hector Garcia-Molina, Ramana Yerneni
In various contexts (e.g., the Internet), the query-processing capabilities of data sources may be limited. Middleware systems based on a mediation architecture are employed to provide powerful query processing interfaces to data sources with limited query capabilities. Such systems also provide integrated views to the data across multiple sources. In this paper, we discuss how the query capabilities of data sources can be described and how mediators support extended query sets over the data sources. We also describe how the set of queries supported by a mediator can be computed from the corresponding sets of queries supported by the data sources integrated by the mediator. We present ways in which content description of data sources can help in extending the set of supported queries. Finally, we describe the challenges posed by diverse source capabilities when supporting robust query processing through replicated sources in the presence of source failures.