Re: Category Types
Date: Fri, 20 Jun 2003 10:46:50 +0100
Message-ID: <bculap$1ofk$1_at_gazette.almaden.ibm.com>
"--CELKO--" <71062.1056_at_compuserve.com> wrote in message
news:c0d87ec0.0306181105.644facc2_at_posting.google.com...
> >> They [modern RDBMS] does not support foreign constraints like
> REFERENCES (tableA or tableB or tableC). <<
>
> You need to know exactly what set of values are referenced.
>
> A reference cannot be one of many different kinds of things.
Agreed, and in Hugh's proposal it does not.
The type of all the attributes in a "distributed key" or a "distributed
foreign key"
are (obviously) required to be the same. In his example they are all of type
(person) Id.
[snip]
> We discussed this "Pendant clause" in the X2H3 Database Standards
> Committee years ago and found problems with circular references and
> domain violations. I do not remember all of the details, but the
> problem was that you could not get a table as a result. One row would
> have more columns than another.
I've no idea what a "Pendant clause" is, but there are no domain violations in "distributed keys" and "distributed foreign keys", and certainly no instances of 'one row with more columns than another' (whatever that might mean).
Regards
Paul Vernon
Business Intelligence, IBM Global Services
Received on Fri Jun 20 2003 - 11:46:50 CEST
