Re: Stupid Database Tricks (was: Little Design Mistakes)

On 22 mai, 16:47, "David Cressey" <cresse..._at_verizon.net> wrote: [Snipped]
> I prefer the title "Stupid Database Tricks". This title is derived from
> David Letterman's "Stupid Pet Tricks". I hope you can see behind the
> levity, and catch some real serious points here.
Yes.
The point here is not about a setting up a hack list but rather to trigger more serious thinking. Manipulation should be in the toolset to fight ignorance. Pointing out misconceptions does not seem to me as teaching a hack. I guess the limit is blurry.

> Sometimes, humor is the only sane response to absurdity.
Nice. That's my quote of the day ;) I would add *sarcasm* to the list.

> A few sample stupid database tricks:
>
> We back up our database every day. But, come to think of it, we have never
> actually tested our backups.
> We found that views run awfully slow, so we forbid views at our shop.
> We found that indexes slow things down, so we don't allow more than one
> index on each table.
> We found that indexes speed things up, so we have an index on every column
> of every table!
I would not put the above in a Design section but on a *Making data administration life less miserable when using direct image systems *. Do you get the idea?

> Data normalization? We don't have time for all that theoretical crap!
> We're working under a deadline here! Talk to us after the version 1 rollout
> is complete.
Bringing back some bad irritating memories from a previous life. At IBM Global Services, I made *Most hated person of the company* at least 3 month in row.(over 450 000 people that's an achievement don't you think?)

For whatever relief that may bring, know that I am in position where people do not take chances to bring that kind of argument against me. All I can say here is that I have the opportunity to practice as a DBA on *very* sensitive systems (Defense). It's good to know that *some* people do not play anymore at computing. (but I am worried about current trends)

> Using our model, you can accomodate any new data that comes along with zero
> changes to the metadata. That way, the DBA can't act as an obstacle to
> progress!
> Our problem is a really simple one, so we store everything in one giant
> table!
>
> You get the idea...
I would not call it an idea. ;) (hope you get the fun)

Regards... Received on Tue May 22 2007 - 19:21:51 CEST