Achieving zero information-loss in a classical database environment
Gautam Bhargava, Shashi K. Gadia · 1989
Abstract. The research in temporal databases has, so far, concentrated on the history of an object as it exists in the real world. Instead, in this paper we view the history of an object as it is recorded in a database. Such a history is obtained by extr lating the outcome of the u$ttes (insert, modify, and Yr elete) made to the object at drscrete instants. Our model su ports two kinds of query-users: the s stem-user, and the coszical-user. P For the classical-user, t E e interface to the database is identical to the usual inter-face in classical snapshot databases. We extend the classical relational model so that a transaction, i.e., an update or a (retrieval) query is recorded in such a way that its effect can be determine d at any time in the future; thus, our model is a zero information-loss model Theorem 1). The logical structure im oeed upon the mode I allows us to give a powerful algebra P or the system-user to query the circumstantial information surrounding updates and queries. In addition, a single execution of a query can be identified with the relation it retrieves; thus, a user can query queries, query queries on queries, ad infinitum. The model represents an application of temporal databases to mainstream databases. It can be used in auditing, and as a foundation for building secure systems. 1. INTFUJDUCTION. An update operation in classical databases is destructive- it not only destroys the environment in which it is exe cuted, but also destroys the environment for queries. After an update is made, only the new database state is available, without even a clue to its past states. A transaction in a database system is either sin update, or a query. The activities in a database system consist of a sequence T of such transactions. We present a model, called the zero injbmatin-loss model, in which no