Re: Checking a relation in BCNF for MVDs
Date: 6 Jan 2003 21:37:12 +0100
Message-ID: <3e19e8f8$1_at_news.uia.ac.be>
In article <c0d87ec0.0301061029.41417738_at_posting.google.com>,
--CELKO-- <71062.1056_at_compuserve.com> wrote:
>>> Let us say there is a relation rel(attr1, attr2, attr3, attr4). It
>is an all-key relation and for that reason we know it is in BCNF. <<
>
>Actually, you know it is in 5NF.
How about the following example. A relation SPJ(S#,P#,J#) with the
candidate key {S#,P#,J#} and the join dependency *({S#,P#},{S#,J#},{P#,J#}).
Two questions:
1. Do you think such a relation can exist?
2. What normal form is it in?
>Chris Date likes to quote a formal proof of this, but I cannot remember the
>details.
Perhaps you remember what I wrote a while ago about good and bad definitions of 5NF?
Guess which definition of 5NF Date presents in his latest seventh edition. :-)
- Jan Hidders