A Measured Evolution of Database Integrity

Autores UPV
Revista Lecture Notes in Computer Science


Inconsistency in large database systems is commonplace and therefore must be controlled in order to not get out of hand. Consistency in database systems is encoded by integrity constraints. Inconsistency thus corresponds to constraint violations. Database system services need to function in spite of extant integrity violations, but inconsistency should not increase beyond control in the course of the evolution of such systems. Evolution is effected by updates that may involve insertions and deletions of relational facts as well as schema updates. We show how to determine the causes of violations. Knowledge about such causes can be used to control inconsistency: an increase of integrity violations by updates can be prevented, while tolerating extant inconsistencies, even if the database schema is altered, and even if the schema is unsatisfiable.