Re: Functional Dependencies > Uniqueness Constraints

From: Marshall <marshall.spight_at_gmail.com>
Date: 4 Sep 2006 12:11:40 -0700
Message-ID: <1157397100.092318.240280_at_m79g2000cwm.googlegroups.com>


Bob Badour wrote:
> Jon Heggland wrote:
>
> > Marshall wrote:
> >
> >>And at the physical level, it seems to me that what you
> >>really want is the FDs themselves. Am I missing something?
> >
> > I would have thought that keys in the form of unique indexes are very
> > important at the physical level. What am I missing?
>
> While I could be wrong, I assumed he was referring to semantic
> optimizations like using the FD's to determine the minimal set of
> constraints the dbms must check for an update to guarantee all constraints.

The specific optimization I was thinking of was in minimizing network traffic. Someone mentioned to me the other day that some product he knew of would optimize transmission of a cross product by sending the unjoined operands and letting the client code assemble them. So the amount of data transmitted would be order m+n instead of order m*n.

It occurred to me to wonder whether this technique could be used more generally than just with cross products. And I believe it can be, but you need to know the FDs to do so.

I would not expect this technique to be useful in very many cases.

Marshall Received on Mon Sep 04 2006 - 21:11:40 CEST

Original text of this message