redundant in r if and only if it's either partly redundant or fully redundant. The paper shows that (a) such a tuple
exists and is partly redundant if and only if R isn't in BCNF (i.e., if and only if R is FD redundant─see Chapter 13);
(b) such a tuple exists and is fully redundant if and only if a tuple forcing JD holds in R (i.e., if and only if R is JD
redundant─again, see Chapter 13). Note, therefore, that “fully redundant” is not a special case of “partly
redundant”; in fact, a tuple can be partly redundant without being fully so or the other way around. Finally, tuple t is
essential in r if and only if it's not redundant in r . If R is in ETNF, then every relation r that's a legitimate value for
R is such that every tuple is essential in r . Note: Our choice of the term essential for use in this context was
influenced by Codd's notion of essentiality, introduced in E. F. Codd and C. J. Date: "Interactive Support for
Nonprogrammers: The Relational and Network Approaches," in Randall J. Rustin (ed.), Proc. ACM SIGMOD
Workshop on Data Description, Access, and Control─Data Models: Data-Structure-Set versus Relational (Ann
Arbor, Michigan, May 1st-3rd, 1974). 3 Briefly, to say some data construct is essential in Codd's sense is to say its
loss would cause a loss of information. As already indicated, every tuple in every relation that's a possible value for
an ETNF relvar is essential in this sense.
3 Republished in C. J. Date, Relational Database: Selected Writings (Addison-Wesley, 1986).