Re: Sixth normal form

From: Brian Selzer <>
Date: Mon, 27 Aug 2007 16:30:18 GMT
Message-ID: <u_CAi.51010$>

"Brian Selzer" <> wrote in message news:qoLzi.21288$
> "David Cressey" <> wrote in message
> news:4JFyi.3591$wr3.2580_at_trndny04...
>> "Brian Selzer" <> wrote in message
>> news:9NDyi.28722$
>>> begin with. I never really came up with a clear definition of what it
>> means
>>> for a schema to have /at least as much/ information, nor did I provide
>>> any
>>> proof that if a relation is in 5NF, then the relationships between the
>>> dependent attributes are due to the fact that the relation is in 5NF, or
>>> that the presence of a pathological relationship between the dependent
>>> attributes indicates that the relation isn't in 5NF.
>> Part of the problem may be that the phrase "at least as much information"
>> suggests some sort of measure of information, but not the information
>> itself. For example, if you have "at least as much money in the bank as
>> I
>> have", it doesn't mean that you have the same money as I do in the bank.
>> I think what you may have meant might be better conveyed by a phrase like
>> "at least all the same information as". But I'm not sure what you did
>> mean,
>> so this is just a guess.
> What I was trying to convey by the phrase /at least as much/ information
> is that the only additional information that should ever appear in an
> instance of the more normalized database schema is exactly that
> information that should be allowed but can't be due to the structure of
> the less normalized database schema. For example, if the FDs A --> B and
> B --> C hold in a relation schema {A, B, C}, then it is not possible to
> insert values for B and C without also inserting a value for A. If it
> should be possible, then there is a structural problem which is due to the
> fact that the MVD B ->-> A | C that holds in {A, B, C} is pathological.
> The problem is that while the decomposition into {A, B} and {B, C} makes
> it possible to insert values for B and C without also requiring a value
> for A, it also permits values to be inserted for A and B without also
> requiring a value for C. The inability to insert a value for A without
> also inserting a value for C is due to the fact that the FD A --> C holds
> in {A, B, C}. Even though it is true that the
> FD A --> C implies the MVD B ->-> A | C, it is only due to the fact that B
> appears in the relation schema that the MVD holds.

Correction: A --> C does not imply B ->->A | C; B --> C does. But the MVD B ->-> A | C and the FD A --> C cannot both hold unless at least one of the FDs B --> A or B --> C also holds.

> Therefore, in order to maintain the functional relationship from A to C,
> it is necessary to add the IND {A,B}[B] in {B,C}[B].
Received on Mon Aug 27 2007 - 18:30:18 CEST

Original text of this message