Re: How to normalize this?

On Thursday, May 9, 2013 11:38:09 AM UTC-7, Erwin wrote:
> All in all, it seems to me like if there is a JD between the attributes of the decomposed schemata, the conventional rule for 5NF should prevail despite the "redundancy" you considered.

The INDs hold on S and R iff JD *{AB,BC} holds on their join. So an adequate SR constraint is SR = SR{A,B} JOIN SR{B,C}. Since a JD holds that is not implied by a key, some single-tuple changes will be invalid (ie this version is subject to update anomolies). Whereas the R+S version is in 5nf.

All the other specifics of your message are just consequences of non-5nf vs 5nf.

On Friday, May 10, 2013 1:29:36 AM UTC-7, Jan Hidders wrote:
> R{A, B} S{B, C} with { R.B -> R.A, S.B -> S.C, R[B] => S[B], S[B] => R[B] }
> which perhaps should be normalized to:
> T'{A, B, C} with { T.B -> T.A, T.B -> S.C }

But now the keys imply the JD (and the INDs) so you don't need to declare/mention it (or them). Presumably it should be normalized to this, we can always decompose to 6nf by separating non-key attributes but we don't.

philip Received on Fri May 10 2013 - 11:03:27 CEST