Re: Relation problem

From: Kristian Damm Jensen <>
Date: Mon, 22 Jan 2001 13:59:31 +0100
Message-ID: <>

Jan Hidders wrote:
> wrote:
> > On 21 Jan 2001 00:26:30 GMT, (Jan
> > Hidders) wrote:
> >
> > >I am afraid that these magic words also don't make the problem go
> > >away. :-)
> >
> > Depends on what the problem really is. See below.
> >
> > >This supertype is a generalization of Individuals and Organizations.
> > >That means that every Party is either an Individual or an Organization.
> >
> > It means that one table, Parties, takes the place of the tables
> > Individual and Organization at their original level of abstraction.
> > Attributes common to individuals and organizations move from those
> > tables into the table Parties.
> >
> > It also means that one FK for the table of party addresses no longer
> > comes from either of two domains (Individuals or Organizations), it
> > comes from just one--the table Parties.
> Of course, but it also means that you will still have two tables Indiv.
> and Organ. that contain the class specific attributes and a foreign key
> to the Parties table. And the reverse of this foreign key is a "split
> foreign key".

Correct. That is inherent in the nature of the data and can not be avoided.

What can be avoided is some of the performance problems. Without the supertype referential integrity from Adresses to Individuals and Organizations can not be maintained using an index. With Parties it can.

Ah, you say, but then you will just move the problem on to maintaing referential integrity between Parties and Individuals/Organizations. But as this will not be updated (!) the is reduced to deletion and insertion, which I believe will happen much more infrequently.

Kristian Damm Jensen              | Feed the hungry. Go to |
Received on Mon Jan 22 2001 - 13:59:31 CET

Original text of this message