Re: Who first (publicly) asserted 3NF is "good enough"?

I was thinking of Ron Fagin's proof that

1. Table is in 3NF
2. every key is one column

The phrase "all keys" was what Chris Date used when he wrote a short piece about this in DATABASE PROGRAMMING & DESIGN back in 1992

The same article also had Zaniolo's definition of 3NF:

T is a table, X is any set of columns of T and C is any single column of T. T is in 3NF iff
for every functional dependency X -> C at least one of the following is true:

1. X contains C -- a trivial dependency like C -> C or ABC -> C
2. X contains a key of T
3. C is contained in a key of T

The nice part is that if you drop #3, you get a definition of BCNF, so you can come up with a good teaching example for 3NF versus BCNF. Received on Tue Sep 19 2006 - 16:31:17 CEST