Re: BCNF

On 1/13/2015 3:34 PM, Stefan Ram wrote:
> Sorry for posting this here, but the general database group
> appears to be missing or empty on my server!
>
> I wonder whether someone knows of a theorem akin to the
> following:
>
> When a database is designed according to an
> entity-relationship diagram which only models
>
> - entity tables with a numerical primary key
> and all other rows being simple (int, double,
> boolean, varchar) properties of the entity,
>
> - 1:n relations between such tables, and
>
> - n:m relations
>
> and the 1:n relations are implemented with foreign keys
> and the n:m relations are implemented with tables with
> two foreign-key columns only
>
> then
>
> that database is in Boyce-Codd normal form
> (or some other similar normal form).
>
> Or, otherwise, a counter-example?
>

From your skimpy description (an example would help), the database looks to be in 3NF form - but the actual data being stored is important, and it's impossible to give a sure response without example tables and data.

It may or may not be in BCNF (3.5NF); this is even more critical on considering actual data and how it is used in the design.

There's a pretty good writeup with examples at http://en.wikipedia.org/wiki/Boyce%E2%80%93Codd_normal_form

I would suggest reading it.

BTW - when I'm designing a database, I always start with sample data and what it means. Actual building tables doesn't start until much later.

-- 
==================
Remove the "x" from my email address
Jerry Stuckle
jstucklex_at_attglobal.net
==================