What Does Boyce-Codd Normal Form Do

Normalization research has concentrated on defining normal forms for database schemas and developing efficient algorithms for attaining these normal forms. It has never been proved that normal forms are good, i.e. that normal forms are beneficial to database users. This paper considers one of the earliest normal forms (Boyce Codd normal form) whose benefits are intuitively understood. We formalize these benefits and attempt to prove that the normal form attains them. Instead we prove the opposite: Boyce-Codd normal form fails to meet its goals except in trivial cases. This counterintuitive result is a consequence of the “universal relation assumption” upon which normalization theory rests. Normalization theory will remain an isolated theoretical area, divorced from database practice, until this assumption is circumvented.