Object-oriented database integrity

There is a growing interest in object-oriented databases for computer aided design, engineering design, office automation, forms, and multimedia applications. Most of these applications have requirements for integrity and many have additional secrecy requirements. Integrity and secrecy mechanisms are best implemented in the database management system so they are uniformly enforced across applications. At present there is not a formal object-oriented data model which is generally agreed upon. The first result of this research is a new, simple formal model of object structure, inheritance, and method invocation for object-oriented databases. In this model, both classes and instances are objects. For better integrity and secrecy, attributes are not accessible outside their respective classes. They must be accessed through methods of the class; this is true even for descendants of a class. Therefore, the accessibility of methods determines the inheritance mechanism. The model supports multiple inheritance. Both the model and its inheritance mechanism are shown to be sound and complete. The second result of this research is a model of integrity for object-oriented database management systems based on first order predicate calculus. The formal object-oriented data model is extended with standard operators (including two aggregation operators). The integrity system is layered over this extended model. The resulting model is shown to handle common state integrity constraints, including domain, key, and structural constraints. This model can enforce value-based functional dependencies from the relational model, both within an object and across object classes. A new relationship structural constraint, reference relationship, is presented which generalizes 1:1, 1:N, and N:M constraints in addition to total versus partial participation. The third result of this research is an approach to integrity and secrecy in object-oriented database systems based on protected groups. A protected group is a set of objects which will only accept messages from one or more interface objects. We show how discretionary access control, data integrity, access integrity, and the Clark-Wilson integrity model can be implemented with this approach. We also discuss how Abrams' Generalized Framework for Access Control may be implemented with this approach.