Term Subsumption Languages (TSLs), a generalization of both semantic networks and frames equipped with a model-theoretic semantics, have shown to be of vital practicability for the representation of knowledge. Recently it has been claimed that the modelling power of TSLs could be decisively enhanced by introducing type constructors as they are used in modern typed programming languages.

In this paper we do a first step into this direction regarding only those type constructors that are commonly used in semantic data models (SDMs). By exploiting the similarity of abstraction mechanisms in SDMs and TSLs we define the integrated representation model IFO+ and demonstrate that the inference mechanisms of both approaches can also be unified. We show in particular that subsumption can be regarded as a stronger version of absolute dominance.

It is claimed that object-oriented databases overcome many of the limitations of the relational data model especially by generalizing the notion of object identification. A clear distinction between objects and values turns out to be essential for the object-oriented approach whereas the relational model is based exclusively on values. Since, however, value uniqueness within scopes is a quite natural constraint for a wide class of applications, identification by value is also of interest for object-oriented datamodels.

Hence, in this paper we concentrate on those classes where the extents are completely representable by values. We formalize some basic concepts of object-oriented databases and show that the finiteness of a database and the existence of finitely representable recursive types are sufficient to decide value-representability.

Another advantage of the relational approach is the existence of structurally determined canonical update operations. We show that this property can be carried over to object-oriented datamodels iff classes are value-representable. Moreover, in this case database consistency with respect to implicitly specified referential and inclusion constraints will be automatically preserved.

In contrast to the relational model methods in OODBs must enforce structurally defined constraints such as inclusion and referential constraints. It has been shown that this is possible for basic generic update operations that are determined by the schema. However, such operations only exist for value-representable classes.

In this paper we generalize this result and show that integrity enforcement is always possible. Given some arbitrary method S and some static or transition constraint I there exists a greatest consistent specialization (GCS) S(I) of S with respect to I. Such a GCS behaves nice in that it is compatible with the conjunction of constraints, inheritance and refinement.

Moreover, it is possible to derive simple representations of GCSs for basic update operations with respect to distinguished classes of explicitly stated static constraints. For the GCS construction of a user-defined operation, however, it is in general not sufficient to replace the involved primitive update operations by their GCSs.