Re: 1 NF

frebe wrote:
> Hi,
> I read an interview with Chris Date (http://www.dbmsmag.com/
> int9410.html) which made me a little bit confused. He claims that
> having an array as a column values doesn't violates 1NF. Is this the
> common opinion at comp.databases.theory too? If yes, how is it
> possible at apply 1NF at all, in that case?
> ...

An example which is maybe not a common use but I like because it's easy for me to remember is the relation that contains the combinations of parts ever shipped in one or more shipments, a set of sets if you will.

If it is a single attribute relation, then it may be hard to display without some device such as a "generated key". Still, the join operator et al could deal with it. To deny its possibility seems tantamount to saying that the relation is not possible in practice unless it has more than one attribute. If certain relations we can imagine are not possible in Codd's RM without an arbitrary number of attributes then I'd think that his theory is broken. I think he said that a relation could have any number of attributes (although he may have excluded the possibility of zero attributes).

But Codd was smarter than I and it was he who coined the term "normal form". Maybe he meant that such a relation has no practical use. If so, I could accept that's what he meant even though I personally wouldn't obey it. It would be like saying that the rest of normalization theory should be applied only when we feel it is practical.

Some pedants would like to re-define it. Or re-number the other normal forms. I'm guessing they find some kind of psychological importance in the number one. I say what for? In any field, including literature, there are lots of terms that have fallen by the wayside. Doesn't bother me to throw the term in the garbage.

p Received on Wed Feb 28 2007 - 23:05:57 CET