Re: implement a referential integrity constraint (long)

From: paul c <toledobythesea_at_oohay.ac>
Date: Wed, 13 May 2009 22:31:39 GMT
Message-ID: <fTHOl.26700$Db2.16078_at_edtnps83>


Gene Wirchenko wrote:
> "Walter Mitty" <wamitty_at_verizon.net> wrote:
>
> [snip]
>

>> However, when I recommend this to you, xyzzy,  I think you will treat my 
>> suggestion with the caveat it deserves.  BTW, I'm sorry you're stuck in a 
>> database that resembles a bunch of twisty little passages, all different.

> ^^^^^^^^^
> Aren't they all the same? Over and over again.
>
> It is hardly that it has never been seen before.
>
> For those who have led a sheltered life:
>
> Announcing a new beginning for Entity Attribute Value DBMSs: EVE.
> Do not put your constraint handling under one roof when you can apply
> it all over on the eaves. Applications are for applying. ... Excuse
> me. I am getting ill. You should, too.
>

The message didn't have much to do with theory, so I wouldn't blame anybody for giving a pragmatic suggestion. If it really is an app that's meant to be sold, there might be other ways to go at it, such as automating the configuration set. That's been done for a long time with lots of products but it seems that many db app producers don't realize they could do it too. I've even seen miniature compilers used once or twice. The fact is, most apps could be compiled, including many 'tables' and this doesn't subvert relational theory.

Of course the real theory problem is that people who design products like Oracle usually wouldn't recognize a constraint theory from a hole in the ground. Hard to blame the small-fry for that but surely Oracle has the dough to spend a few million on a foundation that would go beyond the miniscule efforts they and similar products show - keys, foreign keys and slough the rest off on some half-assed scripting language. Even Date gets mixed up by constraints, suggesting that you can infer results from them whereas you can only infer possible values.

Regarding putting constraints in the dbms, somewhere Date made the incisive comment that nearly everybody got it wrong in the 1960's and 1970's, thinking that logic had to come out of programs when obviously (for other reasons besides constraints), it has to be in the 'front-end' too. The so-called AJAX people persist even today in missing this boat. Received on Thu May 14 2009 - 00:31:39 CEST

Original text of this message