Update Control in Deductive Object Bases1

Deductive object bases are a combination of both deductive and objectoriented databases. This dissertation investigates the problem of update control for such databases. The data model is defined as an extreme case of a deductive database with only one extensional predicate and a multitude of axioms defining the object-oriented abstraction concepts. The advantages are twofold: a distinct conceptual data definition language becomes obsolete, and by a compilation technique the resulting deductive database is more efficient than its pure deductive counterparts. Background and Goals Relational databases became a success in academia and in industry because of their clear theoretical foundations [Codd 70, Reiter 84] and efficient implementations. Deductive databases extended the pure factual data representation in relational databases by a subset of first order predicate calculus, 1 Available as “Änderungskontrolle in deduktiven Objektbanken”, DISKI-19, Infix-Verlag, Sankt Augustin, Germany, 1992. the so-called deductive rules. Deductive databases share the clear semantics with their predecessor but they did not yet become a commercial success. In parallel, object-oriented databases were proposed (e.g., [Maier and Stein 87]). Their data model goes beyond the limitations the relational data model by introducing a richer set of abstractions (classification, specialization, aggregation, and abstract data types). The main disadvantage of object-oriented databases is the lack of declarative component to express simple, i.e., decidable, statements for derivation of data (queries and rules) and consistency checking (integrity constraints). The goal of the dissertation was to find a definition of deductive object bases that literally makes any method for update control in deductive databases applicable without compromising on efficiency.