Re: Importance of the normal forms

Juan Pardillos wrote:
>
>I'd like to know which formal forms are the most important, both from
>a theoric point of view and from a practical point of view. I mean, is
>enough with considering BCNF or we should go on, if possible to get
>4FN or more?.
>
>Which is the usual way of proceeding?.

The usual way is that many people stop at 3NF because they were taught that this the most practical NF and if you go any further things get too difficult, or too inefficient, or "overnormalized" or whatever. What people *should* do is check if they can proceed to BCNF and 4NF and still remain dependency preserving. Sometimes this will make certain queries harder to compute, especially those that have to join again the tables that were split in the normalization process, but whether this is significat may depend on a lot of details and on the other hand not normalizing can maked queries harder if they do not need the join. Moreover, not being in 4NF can cause lots of redundancy and make your tables much bigger then necessary.

>As an student, I've been only taught until 4FN, but I think there are
>some more. I'd like to know if they are really important and/or easy
>to understand. Any link is greatly appreciated.

There's only one more, 5NF (also called PJ/NF or PJNF or the project-join normal form), and it has in fact been proven that this is the highest normal form that can be reached by splitting relations by decomposing them into projections. For a small example see:

http://www.gslis.utexas.edu/~l384k11w/normover.html

It's one of the hardest normal forms to understand and in practice it is not easy to spot the special type of dependencies, i.e., the join dependencies, that are used to check if you are in 5NF. In fact, it is a public secret that there are a few text books out there, some of them by well-known authors, that don't define it correctly.

• Jan Hidders