On Fri, 25 Sep 1998, Vadim Mikheev wrote:
> I don't like implementation of referential integrity constraints
> in contrib/spi/refint.* and implementation of ... unique indices:
> backend checks constraints when tuple is changed.
>
> Oracle checks constraints after entire _statement_ is done.
> Do you remember this:
>
> vac=> create table x (a int unique);
> NOTICE: CREATE TABLE/UNIQUE will create implicit index x_a_key for table x
> CREATE
> vac=> insert into x values (1);
> INSERT 143852 1
> vac=> insert into x values (2);
> INSERT 143853 1
> vac=> update x set a = a + 1;
> ERROR: Cannot insert a duplicate key into a unique index
>
> This is simplest case. Keeping in mind triggers
> I would like to see Oracle behaviour some day and so
> I decided to do not implement FOREIGN keys in 6.4 -:((
Sounds reasonable to me...all things considered, I think that v6.4
has turned out, once more, to be one giant leap forward :)
> Sorry.
Don't be...I think if Bruce had his way, we'd be funding a cloning
project just so that we could get a dozen of you :) Its nice to be
reminded once in a while that you are "only human" *grin*