Re: Testing relational databases

From: Chris Smith <cdsmith_at_twu.net>
Date: Mon, 10 Jul 2006 11:01:59 -0600
Message-ID: <MPG.1f1c35e83d04d493989776_at_news.altopia.net>


S Perryman <a_at_a.net> wrote:
> "Marshall" <marshall.spight_at_gmail.com> wrote ...
>
> >S Perryman wrote:
> >> Design By Contract ??
> >> Specification-directed testing ??
> >> Design For Testability ??
>
> >> All techniques that by themselves or together will give you for a large
> >> majority of systems exactly what "we want" .
>
> > I don't follow. What are you trying to say?
>
> The techniques (see above) to deliver what "we" want are already here. :-)
> And they work very well indeed.

It's unclear what you mean by "design by contract", and specifically, whether you've got in mind tools that verify the contract universally or existentially. Certainly neither of the other techniques you listed are sufficient to provide universal proofs of program behavior for computers of any substantial size. Testing all possible computer states for a computer with even 1 MB of memory requires an amount of time that dwarfs the entire length of the universe to date.

I can tell you rather conclusively that static analysis tools perform some limited tasks rather well, but they are no where near proving correctness of arbitrary programs from arbitrary specs.

-- 
Chris Smith - Lead Software Developer / Technical Trainer
MindIQ Corporation
Received on Mon Jul 10 2006 - 19:01:59 CEST

Original text of this message