Archives of the TeradataForum

Message Posted: Wed, 21 Jan 2004 @ 14:13:31 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

Clay,

ANSI do not specify how index DDL should look like so it is all up to the RDBMS vendor, and, just like you, I can only be guessing here but I thought there could be some performance considerations taken into account. If we allow UNIQUE in PI or SI definitions of JI, we would have to be checking for such uniqueness conditions *before* the data is modified in the base tables.

I think (and this is only my guess) this uniqueness check on JI must have been done before the data is actually modified in the base tables because otherwise we would get in trouble with some general referential and transactional integrity rules, which ANSI do specify. If my guess is correct, it would mean that UNIQUE clauses on JI would always create some extra performance overhead with any DML on the base tables (whether it will on not violate later the uniqueness condition on JI), and that's probably why Teradata designers decided to avoid them. There are surely enough means to take care about uniqueness while designing the base tables themselves.

Again, this is only a guess, and I would also be interested to hear from somebody at NCR on this topic, if they have time to take this question.

Regards,

Victor

	<Prev	Next>		<<First	<Prev	Next>	Last>>