Integrating Dynamically-Fetched External Information into a DBMS for Semistructured Data

We describe the external data manager component of the Lore database system for semistructured data. Lore's external data manager enables dynamic retrieval and integration of data from arbitrary, heterogeneous external sources during query processing. The distinction between Lore-resident and external data is invisible to the user. We introduce a flexible notion of arguments that limits the amount of data fetched from an external source, and we have incorporated optimizations to reduce the number of calls to an external source. 1 Introduction Most research on querying distributed, heterogeneous, information sources takes one of two approaches. The first approach is data warehousing, where information is brought into a single repository for querying and analysis [Wid95, LW95]. Drawbacks with this approach are that changes to the original data may not be visible immediately, and that all information of interest must be replicated in advance. The second approach is to retrieve informatio..