Re: Dreaming About Redesigning SQL

From: andrewst <member14183_at_dbforums.com>
Date: Thu, 23 Oct 2003 08:12:45 -0400
Message-ID: <3514739.1066911165_at_dbforums.com>


Originally posted by Mike Preece

> andrewst <member14183_at_dbforums.com> wrote in message
> news:<3505221.1066738677_at_dbforums.com>...

> > And if you are just one of 100 developers working

> > on a project, your project management can just assume that you
> all do

> > the right thing (use the standard function)?

>

> Never assume! If you've read my other posts in this and other forums

> you'll have seen that I'm very much against people making assumptions.

> No. It must, of course, be a QA function to ensure the correct

> procedures have been followed before code is "released" or "put live".

>

> > Or will every line of

> > every programmer's code be subject to scrutiny by someone who
> infallibly

> > spots EVERY single deviation from the business rules? Sure,
> that is an

> > option. There may be 200 modules that have to write data to a
> core

> > table. You can eyeball each one to ensure it does it correctly,
> or you

> > can guarantee that they each do it correctly (or not at all) by
> ensuring

> > that there is no possible way to bypass the rules. For a
> small

> > database/application developed by one person, you may be right;
> for a

> > large database with teams of developers writing code - I don't
> think so.

>

> Funnily enough, I wrote the IO functions for a "Pick" (actually

> UniData) based system for the UK government. Quite a large and

> important application. And no - there is no problem.

>

> Cheers

> Mike.

OK, so you wrote all the IO functions, and presumably every developer is expected to use them, and a QA process ensures they do? That's good.

But tell me: how is that better than having an architecture where your IO functions are the ONLY way to touch the data, rather than relying on QA procedures? Or than having declarative integrity IN the database, so that even if they don't call your functions they can't mess anything up?

Analogy: is it better to have the innards of every developer's PC exposed, and rules telling them not to touch the inside, or is it better to have a case on it that prevents (at least accidental) poking around?

--
Posted via http://dbforums.com
Received on Thu Oct 23 2003 - 14:12:45 CEST

Original text of this message