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

On Wed, 29 Jun 2005 13:51:47 GMT, "David Cressey" <david.cressey_at_earthlink.net> wrote:

[snip]
>"In the ER model, relationships are identified but not implemented."

In the relational model, too.

>The idea here is that lines on paper (or elements in a model) specify
>relationships, give them names, connect them with entities, and describe
>them. But they don't provide a specific data structure for "realizing"
>those relationships at retrieval time.
>
>"Foreign keys", implement relationships. They aren't the only possible
>implementation, but they are by far the most prevalent in today's practice.

At the logical level, *nothing* is implemented. The logical level is, umm, logical, not physical.

A foreign key is just one of many kinds of *logical* constraints. Implementation of logical constraints happens at the physical level.

In modern SQL platforms, you can usually implement logical foreign key constraints with a FOREIGN KEY ... REFERENCES ... clause in DDL. But all the logical constraints apply even if you don't use SQL. In fact, all the logical constraints apply *even if you don't use computers.*

For a practical example, pick a logical model that's been implemented in SQL and moved to production. Then implement it again in Lotus Notes/Domino.

-- 
Mike Sherrill