TECHNICAL CORRESPONDENCE.

The article takes a fresh look at first normal form (1NF). A fresh look at E.F.Codd's original article on relational databases suggests a new, non-trivial definition of 1NF. Let R be a relation scheme and D the set of dependencies applicable to R, including multivalent dependencies. Y is a simple component of R if Y is a nonempty subset of the attributes of R, and there is a multivalued dependency X→→Y, where XY does not include all the attributes of R, and there is no non-empty subset Z of Y such that X→Y. The last clause is necessary to guarantee that there are at least two tuples that agree on X but do not agree on Y, thus giving Y complexity. The definition of nonsimple component seems to agree closely with Codd's intuition. The improved definition of 1NF is simple: R is in 1NF iff R has no non-simple components. If it is accepted that the problem of nonsimplicity is the problem of multivalued dependency without functional dependency, then one would conjecture that if the problem of nonsimplicity is taken care of, then the only remaining problem is that of functional dependency.