Websrcmed: a Mediator for Scaling up to Multiple Web Accessible Sources (websources) Universidad Simm on Bol Var Caracas Venezuela

The rapid growth of the Internet and Intranets has dramatically increased the number of WebSources. These sources are typically characterized by a limited query capability compared to a relational data source. They may also be incomplete, and there may be redundancy among them. These features increase the challenge of scaling to answer queries in a mediator-wrapper architecture with a large number of WebSources, and they impact the mediator tasks of capability based rewriting (CBR) and query optimization. We propose a mediator-wrapper based architecture for WebSources, WebSrcMed. Using query capability, content and coverage (completeness) descriptions of WebSources, we develop techniques for CBR that maintain the following properties: rst, we ensure that WebSrcMed only accepts queries for which it can guarantee that there is at least one WebSource that can evaluate this query { this is a safe query, for which there is a safe plan. For each safe query, we identify when the largest possible subquery can be submitted to a WebSource, depending on its capabilities, and this feature results in a maximal plan. Finally, if there is redundancy of complete descriptions of WebSources, we identify this redundancy in computing the answer and produce non-redundant plans. To maintain these properties, WebSrcMed exports a single interface which reeects the descriptions (capability, content and coverage) of multiple WebSources, for some domain. This interface is deened so that WebSrcMed can use it to accept only those queries for which there is a safe plan. To provide possibly maximal and non-redundant plans, WebSources and their descriptions are partitioned into equivalence classes. These equivalence classes are partially ordered within a lattice of complete WebSources that are redundant, and a lattice of possibly incomplete WebSources. WebSrcMed uses these lattices to identify relevant WebSources and to identify a set of alternate WebSources. The lattices also provide a description for each relevant WebSource, so that the WebSrcMed can produce a possibly maximal plan.