Functional Dependencies and Incomplete Information

Functional dependencies play an important role inrelational database design. They are defined in the contextof a single relation which at all times must containtuples with non-null entries. In this paper we examinean extension of the functional dependency interpretationto handle null values, that is, entries in tuples thatrepresent incomplete information in a relational database.A complete axiomatization of inference rules forextended functional dependencies is also presented.Only after having such results is it possible to talk aboutdecompositions and normalization theory in a context ofincomplete information. Finally, we show that there areseveral practical advantages in using nulls and a weakernotion of constraint satisfiability