Just-In-Time Databases and the World-Wide Web

Abstract

It is not always possible or practical to store very large data sets in a relational database. For example, in a rapidly changing large-scale environment with distributed control, such as the World-Wide Web, a strict relational approach is not feasible. Nevertheless, it is desirable for a user to be able to make SQL queries on the entire data set, because SQL is well known, supported, and understood. We introduce "just-in-time databases", which allow the user to query the entire data set as though it were in a relational database. The underlying engine brings data into the relational database only when it is required by user queries. We describe how "just-in-time databases" are implemented, using the WorldWide Web as an example.