Re: Stupid Database Tricks
Date: Wed, 23 May 2007 15:35:27 GMT
Message-ID: <3bZ4i.209931$DE1.169732_at_pd7urf2no>
Roy Hann wrote:
> "Marshall" <marshall.spight_at_gmail.com> wrote in message
> news:1179892760.282700.315160_at_q66g2000hsg.googlegroups.com...
>
>>On May 22, 1:48 pm, Bob Badour <bbad..._at_pei.sympatico.ca> wrote: >> >>>Every table will have a numeric id column, and this column will be the >>>primary key. >> >>I don't know why, exactly, since it's not like it's the worst one >>mentioned, but that one DRIVES ME CRAZY! >> >>I hear people say this, and I want to say oh, I see: you're >>just the stupidest freaking idiot ever, is that it? Of course >>instead I say something about that being contraindicated.
>
>
> I did a database design briefing for a regional outpost of one of the big
> three consultancies just the week before last, and I actually spent quite a
> lot of time preparing them to be receptive to my arguments debunking the
> need for a numeric ID column in every table.
>
> Well damn me; some B***TARD had got there before me and implanted the idea
> that "sequentially" numbering the rows in a table is not just OK but is in
> fact "best practice" (that was his exact phrase)! And once he'd put that
> completely bogus label on it (with no justification that they could remember
> or explain to me), then as far as they were concerned it would be reckless,
> negligent, and nothing short of unprofessional to deviate.
>
> No amount of explaining or probing or proposing problems with it would shake
> them from the idea that a numeric ID column is the hallmark of a good
> design. They were eager to seem educated and well-informed, and taking a
> positive step (conspicuously adding a column) was, to them, a more obvious
> way of demonstrating their expertise than the passive (and invisible) step
> of not adding one. Psychology. Humph.
>
> I was gutted. I'd gone in there prepared to do battle against people
> advocating object-relational mapping frameworks that insist on row numbers,
> and it still wasn't enough.
>
> Roy
>
>
The insidious aspect of this I used to notice is that there is usually some marketeer running the outpost who's smart enough to know that such "practice" has a good chance of increasing billings. I saw one multi-user app where the next shipment number was used as a reference and so needed to be known at commit time, convoy phenomenon quickly took effect and app crawled to a halt soon after testers logged on and the marketeers saw the opportunity to sell additional stress testing to the project, more consultants too, not to mention the re-write.
I really do believe that "successful" "big three" outposts are mindful, at least subconsciously, that ignoring elemental features usually creates profitable complications.
p Received on Wed May 23 2007 - 17:35:27 CEST