Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!iuvax!cica!tut.cis.ohio-state.edu!gem.mps.ohio-state.edu!ginosko!uunet!mcvax!cernvax!ethz!marti
From: marti@ethz.UUCP (Robert Marti)
Newsgroups: comp.databases
Subject: Re: A few words on the "normalization"
Message-ID: <1707@ethz.UUCP>
Date: 5 Aug 89 09:45:17 GMT
References: <36270004@hpindwa.HP.COM> <36270005@hpindwa.HP.COM>
Organization: ETH Zuerich
Lines: 26

In article <36270005@hpindwa.HP.COM>, donovan@hpindwa.HP.COM (Donovan Hsieh)
writes in repsonse to some earlier message of mine:

> My point is, it is difficult to formulate all possible MVDs for a given 
> relational database scheme in the real world applications. Certainly, finding 
> some or several of the MVDs are easier. If the designer choose to use the
> synthese algorithm to normalize the database scheme, incomplete set of FDs &
> MVDs can lead to incorrect result.

I see your point, and it's certainly valid.  However -- as I am sure you
know -- a designer does not have to specify ALL functional and multi-
valued dependencies to achieve correct third or fourth normal form.
All she needs to do is provide a superset of an FD cover (or superset of
a dependency basis in the case of MVDs) from which all FDs (MVDs) can be
inferred.  For real world examples, finding these FDs and MVDs is
essentially the same thing as finding the relationships between different
entity types with their association cardinalities in an ER model.
In either case, you get an incorrect model of the world if the designer
overlooks some relationships between attributes (in the relational model)
or entity types (in the ER model).

-- 
Robert Marti                      Phone:      +41 1 256 52 36
Institut fur Informationssysteme
ETH-Zentrum                       CSNET/ARPA: marti%inf.ethz.ch@relay.cs.net
CH-8092 Zurich, Switzerland       UUCP:       ...uunet!mcvax!ethz!marti