Re: Normalization

> >Your question got me curious, however, so I called around to a couple of
> >folks who would know. The responses can be summarized as: "un-important",
> >"trivial", and "splitting hairs".
>
> I tried a couple of minutes to prove that PJNF implies 5NF and vice versa,
> but I didn't find a proof of either implication. So, I don't think it is

While I am not able to formally offer a proof (not yet, anyway) I'm told by the above folks that PJNF is a stronger form of 5NF, and thus, PJNF implies 5NF.

> very trivial :-). And I also think that it is important because it shows
> that normalization is actually trying to achieve two goals at once: one is
> removing repeating groups, and the other is to organize the tables in such
 a
> way that only key constraints need to be checked. It is easy to see that
 the
> 5NF achieves the first goal and that the PJNF attempts to achieve the
 second
> goal. So the question that we are talking about is the question if these
> goals are conflicting or not. Wich is, in my opinion, an important
question.

5NF and above are more dependent on underlying business rules inherant to the data values than the values themselves. In other words, they may or may not contain "repeating data", depending on your Universe of Discourse (UoD). They only apply to all key tables of 3 or more columns. The underlying business rules determine if certain combinations are allowed or not, and depending on this outcome, the tables could be further split.

Frankly, I find (what I call) "manual normalization" to be a pain. I prefer to perform normalization by the gathering of business rules and the construciton of the conceptual model (ala ORM). I then get a normalized schema for free. For example, I seldom look at a legacy schema to determine if it is normalized or not (except for obvious violations, I suppose). Rather, I perform my analysis techniques and map the results to a logical schema -- and then compare that new schema to the old.

> >The paper Halpin cited is: Orlowska, M. & Zhang, Y. 1992, 'Understanding
> >fifth normal form (5NF)', Australian Computer Science Communications,
 vol.
> >14, pp. 631-9
>
> Ok. Thank you for the reference, I will see if I can find it.

I am doing the same. If you manage to get a copy, please let me know (scotb_at_inconcept.com) -- thanks in advance.

Scot.

Principal Consultant, InConcept, Inc.

     http://www.inconcept.com

Editor, The Journal of Conceptual Modeling

     http://www.inconcept.com/JCM Received on Sun Aug 06 2000 - 00:00:00 CEST