Re: Normalization question
Date: 26 Jun 2002 22:17:43 +0200
Message-ID: <3d1a2167$1_at_news.uia.ac.be>
In article <a386bae7.0206261051.6c87a148_at_posting.google.com>,
leebert <leebertarian_at_yahoo.com> wrote:
>
>Old Data Model (key is ID + Year + Month) :
>
>New Data Model (key is ID + Begin_dt + End_dt, requires trigger for
>complete enforcement)
Actually you have two smaller candidate keys: ID + Begin_dt and ID + End_dt. But even if you enforce both you will still need triggers to prevent overlapping periods for the same ID.
>I know if I went back to 1st principles and read some weighty tome on
>normalization I'd find an answer, but I'm taking a shot & see what
>more experienced folks have to say, save some time narrowing down
>what's best practice.
Classical normalization doesn't really cover this type of redundancy. It only covers redundancy that can be solved by splitting a table into smaller projections. The old and the new data model could very well be in 5NF.
- Jan Hidders