Re: Does Codd's view of a relational database differ from that ofDate&Darwin?[M.Gittens]

From: David Cressey <david.cressey_at_earthlink.net>
Date: Mon, 04 Jul 2005 01:38:37 GMT
Message-ID: <xq0ye.3264$aY6.3017_at_newsread1.news.atl.earthlink.net>


"Marshall Spight" <marshall.spight_at_gmail.com> wrote in message news:1120237546.473797.242280_at_f14g2000cwb.googlegroups.com...
> Right. I think that's the part of what Jan's saying that I'm attracted
> to; that we might raise the level of abstraction. It would be nice
> to just declare the relationships and their cardinality, and not have
> to teach everyone the way foreign key relationships work. I know that
> sounds like of lame when you say it but it's a very real world
> consideration.
>
> However, if doing so means we have to reintroduce the concepts of
> identity or rowids back, then I'd say it wasn't worth it.

What is abstraction? What is it for?

Abstraction is the intentional omission of some detail in order to clarify the detail that is NOT omitted.

Abstraction is a two edged sword. An more abstract model is ALWAYS less "true to reality" (verosimil in Spanish) than a less abstract model. The ultimate description of reality is reality itself, the universe. Everyone experiences it, but no one understands it.

OK, let me get off the hifalutin phrases and back to the conversation.

An ER model identifies relationships, but doesn't specifiy how the relationships will be represented inside the database.

An R model does specify how relationships will be represented: by means of foreign keys. But just saying that implies that entities can be referenced by means of key data. In other words, that Identity is an attribute, or a compound made up of attributes. An ER model is easier to communicate to some people than an R model precisely because it doesn't make that assumption.

An ER model is also one step further away from implementation for the same reason. Received on Mon Jul 04 2005 - 03:38:37 CEST

Original text of this message