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

>> .. one so frequently sees books and articles in which it is asserted that 3NF is "good enough" .. does anyone know if there was a single authoritative writer (long ago) who expressed this foolish idea? <<

I am not sure the source of the myth, I have a feeling it goes back to the early forms of ER diagrams which did not have much in the way of showing compound keys or "tricky" relationships. A lot of the early ER tools could only go to 3NF, so maybe that was where it started.

And, as Bob pointed out, for a lot of commercial apps, there is an industry standard or internally defined key that you have to use (UPC, EAN, VIN, ISBN, etc.), so most of the tables have a simple key. Also, if a table is all key, then it is in 5NF.

When I have to teach a class, I pull out a (buyer, seller, lender) table that demonstrates JPNF problems and a (classroom,course, teacher, period) table where any three columns are a key for overlapping UNIQUE constraints. Received on Mon Sep 18 2006 - 23:55:36 CEST