CO-RELATIONSHIPS, LEVELS OF SIGNIFICANCE AND THE SOURCE OF THE CONNECTION TRAPIN RELATIONAL DATA BASES

Abstract

We propose to attribute one or more precisely defined levels of semantic significance to every co-relationship in a relational data base. We also propose a precise definition of a relationship, and define a co-relationship as one of the two non trivial primitive relationship types, that is, a co-relationship occurs between two relations when a relationship relation can be formed with a single join on join attributes of the two relations that are neither primary nor candidate key attributes. It is shown that a co-relationship will give rise to a connection trap when users assume either a level of significance for the co-relationship that is too high, or which is correct but not supported by the data in the data base. Furthermore, anticipation by data base designers of user assumptions about levels of significance for co-relationships, and consequent data base modifications to avoid connection traps, will be very much a matter of judgement of user behavioral probabilities, thus rendering both automated data base design algorithms and universal relation generation algorithms prone to failure.