Path: utzoo!attcan!utgpu!cunews!cognos!colinm
From: colinm@cognos.UUCP (Colin Moden)
Newsgroups: comp.databases
Subject: Re: Entity Models and BCNF
Message-ID: <9160@cognos.UUCP>
Date: 21 Dec 90 10:04:28 GMT
References: <9129@cognos.UUCP> <34942@netnews.upenn.edu>
Reply-To: colinm@cognos.UUCP (Colin Moden)
Organization: Cognos Inc., Ottawa, Canada
Lines: 68



Aaron, I think you and I are on the same side. 
I too believe that Entity Modelling can be used at the "Logical" level
as well as the "Conceptual", and can provide a full replacement for
Normalization where Entities & Relations correspond 1-1.

Anyway you asked for a real example, so here is one from a customer in
Belgium.  (The context of this was that I was teaching them to Model
before programming).

The system is Car insurance and they came up with the Entities

Policy Types (Key : Policy type No)

Policy (Key : License plate, Policy type No, Start_Date)

Vehicle (Key : License plate)

Owner (Key : Owner Reference No)


I suspected there was more to it than that and asked what the Attributes
of Vehicle were, they gave.

License Plate,
Manufacturer,
Model,
Engine Capacity,
Date of Manufacture,
Car Body type,
Insurance group,
...

By using normalization I was able to show them that they had missed out
entities like : Car manufacturer, Model, Insurance Group

Similar work on Policies identified : Claims, Payments, Quotes, Receipts etc


Of course a good entity modeller would have found these anyway, but few of us are
perfect, and Entity Modelling has no way of checking when you are Done.
By listing attributes and dependencies & normalizing you do get an extra
viewpoint and a check that you are complete.

Of course if you miss attributes or dependencies nothing will save you, but
you are at least internally complete & consistent.


Part of my job is to train people to design systems, by using this combination
of Entity Modelling and Normalization people can check their own work and learn 
from it. Initally people spend their time 50% on each, and find 75% of
their  entities when Normlizing. After a couple of years the Normalization checks 
become almost subconscious and they spend 90% of the time on the ERM, 
5% writing down the attributes and 5% checking that it is fully normalized
with a 95% success rate. 

But it's still a useful check and the addition of attributes at this stage is
the next step anyway.

Colin Moden


-- 
Colin Moden               Voice: +44 344 486668              Westerly Point
Cognos Incorporated       FAX: +44 344 485124                Market St 
uucp: colinm@cognos.uucp || uunet!mitel!cunews!cognos!colinm Brackell, Berks
                                                             UK  RG12 1QB