Re: Normalizing the ER model

Greg Boland wrote:
>
>"K.Y. Fung" <kyfung_at_hotmail.com> wrote in message
>news:c22bb40d.0211211822.2bb16043_at_posting.google.com...
>> hidders_at_hcoss.uia.ac.be (Jan Hidders) wrote in message
>news:<3dda1122$1_at_news.uia.ac.be>...
>> > In article <6278687.0211181428.75890385_at_posting.google.com>,
>> > Juan Pardillos <sicotom_at_eresmas.com> wrote:
>> > >
>> > >is it possible to apply the normalization process directly to the ER
>> > >model instead of apply it to the relational model?
>> >
>> > Yes, you can interpret a relationship as a relation and then apply the
>> > usual normalization rules. The same holds for the entity types but you
>> > have to take care of where the connections with the relationships go if
>> > you split then entity.
>>
>> I am just a beginner studying the process of normalization in OU Hong
>> Kong. I don't understand how normalization can apply to ER model.
>>
>> Correct me if I were wrong. ER model consists of Entity diagram, Entity
>> type and its headings, Constraints and Assumptions. But Entity type only
>> have an identifier without any keys. Entity type only shows there is a
>> need to record information, and we draw a relation between entities base
>> on the requirement we gather from user. But neither the entity heading
>> nor the relation tell us how information in one entity is related with
>> information in another entity. We can only infer it in the relational
>> model when primary key / foreign key mechanism comes in. So how can
>> normalization apply to ER model?
>
>King Yin Fung has it right. . Normalization is applied to attributes, not
>entities.

Wrong. Normalization is not applied to attributes (unless they are relation-valued) but to relations. And since entities can be seen as special cases of relations normalization also applies to them.

>Much of what falls out of normalization are new entities. As a matter of
>fact IDEF requires only identifiers for a "complete model" at the
>conceptual level.

Sure. So?

>Normalization, IMHO, removes update anomalies. I usually only go to 3rd
>normal form.

If you have the chance you should also visit BCNF, 4NF, PJNF and DKNF. I hear the beer is great there. :-)

>What ER modeling provides for me is migration of key/foreign key from on
>entity to another. But then, still, you must go through a final
>normalization .

Really? Can you give an example of an ER diagram that is in some normal form and for which the straightforward mapping to the relational model is not in the same normal form?

• Jan Hidders