Re: A Question On Many-To-Many Linking Table(s)

"--CELKO--" <71062.1056_at_compuserve.com> wrote in message news:c0d87ec0.0209221923.42b0561b_at_posting.google.com...
> aleatory_at_hotmail.com (aleatory) wrote in message
news:<a68a4ee0.0209220745.4442210e_at_posting.google.com>...
> > Hi all,
> >> According to some of the database books I have, they say that
> many-to-many implies a linking table(s). <<
>
> That is a terrible term; tables are either entities or relationships
> and not both.

I tend to not agree with the statement. A table *is not* either an entity or a relationship. A table is a table, though it might, just as a relation would, act as some implementation representation of an entity.

Some individuals, most of whom I tend to agree with, assert that a relationship is nothing more than a special type of entity. Given this interpretation, a table could implement the equivalent of _both_ an entity and a relationship (in contrast to Celko's assertion of either/or exclusiveness). However, there are often cases where it does not map directly to either. Tell me, is a relationship married-to with the attributes husband and wife any different than the conceptual entity, marriage, with the same attributes?

Also, an ER model does not necessarily imply a one-to-one mapping when entities and relationships are implemented as either relations or as tables. Structures can change and evolve so that the eventual implementation looks nothing like the original entities and relationships when originally modelled, especially when considering multi-valued attributes, generalization, specialization, and the normalization process.

Doesn't it make sense to maintain a clear distinction between implementation models and those models that represent objects, concepts, idea's, etc. in the user's mind at the time requirements are being gathered, given that a mapping between the levels should be dynamic and well-maintained?

Dan Guntermann Received on Mon Sep 23 2002 - 20:49:37 CEST