A universal relation database system implemented via the network model

We describe a network schema design algorithm that uses a well-designed relational schema as input. The resulting network database handles incomplete information, since we use only the modified foreign-key constraint and not the universal instance assumption. We believe that the representative instance is a correct representation of information stored in the database. We show how to map the representative instance to the network database. We give two algorithms for translating a relational query to a network application program. The first algorithm is confined to relational queries that express selections and projections over the universal relation scheme. The second algorithm includes all relational queries over select, project and (natural) join. Optimization techniques are discussed for both types of translation.