OK-- back on topic...

I've reviewed some of Date's recent writings on 1NF and have concluded that the table from the article in fact does violate 1NF, but not for the reasons others have mentioned.

I came close when I mentioned that since there is exactly 1 value at each row/column intersection for the strDescription column, the table was in 1NF (this relies on the assumption that the picture in the article is a representative sample of the data).

However, I ignored the strCode column, which allows NULLs. Since a NULL is not a value, the column does not have exactly 1 value in that column for every row. So not only can you not have more than one value (of the particular domain) for the column in a row, you can also not have less than one. ("Exactly one"). Therefore, the table isn't in 1NF.

In the SQL world, we've gotten used to accepting NULL, though, so it doesn't bother us to talk of tables being in 1st (and higher) normal forms even when they allow NULLs in some of their columns.

To close, this quote on 1NF is from "An Introduction To Database Systems (7th Edition)", Chapter 11, section 3, by Date:

"First normal form: A relvar is in 1NF if and only if, in every legal value of that relvar, every tuple contains exactly one value for each attribute."

TroyK