On Fri, Feb 20, 2009 at 04:51:33PM +0100, Ivan Sergio Borgonovo wrote:
> What I find a bit annoying is politely deal with the error once it
> is reported back to the application *and* connection and *bandwidth*
> costs of moving clearly wrong data back and forward.
This sounds a bit like premature optimization to me; I don't think many
people worry about optimizing the failure code paths. I know I prefer
to make sure that things go quickly when they're working. If you're
worried about someone performing a DOS attack on a failure then you'd
want to optimize it, but surely you'd want the checks early in the
application code.
> If you've a good mapping between pg types and the application
> language/library types it becomes easier to keep in sync those
> checks otherwise it is a really boring job and DB checks becomes just
> one more security net to maintain.
It does, but constraints like that aren't going to be changing to
regularly are they?
--
Sam http://samason.me.uk/