Abstract

Many multilevel data models have been proposed, and different models have different merits. In this paper, we unify many of these merits to build a new multilevel data model as a natural extension of the traditional relational data model. We describe our new model called the Multilevel Relational (MLR) data model for multilevel relations with element-level labeling. The MLR model builds upon prior work of numerous authors in this area, and integrates ideas from a number of sources. The new data-based semantics is given to the MLR data model which combines ideas from SeaView, belief-based semantics and LDV model, and has the advantages of both eliminating ambiguity and retaining upward information flow. The model is simple, unambiguous and powerful. It has five integrity properties and five operation statements for manipulating multilevel relations. In order to support this integration, several new concepts are introduced and many old ones are redened. A central contribution of this paper is proofs of soundness, completeness and security of the MLR data model. These proofs respectively show that any of the provided operation statements can keep database state legal (i.e., satisfying all integrity properties), every legal database state can be constructed, and the MLR data model is non-interfering. This is the first time that such properties have been formally proved for a multilevel relational data model. The expressive power of the MLR model is also discussed in this paper, and is compared with several other models.

Citations