Treffer: Functional Dependencies for Object Databases: Motivation and Axiomatization

Object identification by abstract identifiers should be considered as a modeling and not as a database concept. This means that object identifiers are not appropriate for the access to specific objects using a database language. In this paper we discuss how the relational concept of a functional dependency can be adapted to object databases in order to get more convenient ways of accessing objects. Graph based object functional dependencies are proposed as a means to specify constraints between attributes and object types of an object schema. Value based identification criteria can be defined using a special type of object functional dependencies. Different definitions of satisfaction are given for these constraints, based on a so-called validation relation, and their relationships are investigated. These definitions are related to different forms of identification. Using the strongest notion of satisfaction, inference rules for the derivation of new dependencies are discussed with emp.