Re: Declaring Unenforced Constraints

From: Laconic2 <laconic2_at_comcast.net>
Date: Wed, 17 Nov 2004 14:47:52 -0500
Message-ID: <O4qdnclnDvR0NwbcRVn-ow_at_comcast.com>


"Kenneth Downs" <firstinit.lastname_at_lastnameplusfam.net> wrote in message news:cv8s62-i7p.ln1_at_pluto.downsfam.net...

> Agreed, I spoke to broadly in earlier post. A mailing list is something
> that you can safely bring offline for a complete load, while an OLTP
> database you could not.

OK, I'll confess. When I originally started this topic, I DID have a hidden agenda. But it wasn't intended to trip up anyone in this NG. Here's the hidden agenda:

The people who don't want constraints enforced in databases typically offer these objections.
Constraints slow things down.
Constraints prevent the user from storing what needs to be stored, if the constraints are too constraining.

The unenforced constraint answers both of these objections:

It won't slow things down until you test it. And sloppy data can be put into your database any time you like.

Later on, you can find the sloppy data, and decided whether to fix the sloppy data, or fix the constraints, or fix both.

And, you can even find out HOW MUCH it costs to enforce the constraint all the time, by altering the enforcement of the constraint.

I think this would provide a road to redemption for those managers who care more about performance than about correct data. And that's the hidden agenda. Received on Wed Nov 17 2004 - 20:47:52 CET

Original text of this message