Re: looking for a term

From: mAsterdam <mAsterdam_at_vrijdag.org>
Date: Sun, 10 Jul 2005 10:47:50 +0200
Message-ID: <42d0e0ba$0$6214$e4fe514c@news.xs4all.nl>

paul c wrote:
> mAsterdam wrote:

>> paul c wrote:
>> ...
>> How would they differ in meaning?

>
> perhaps i shouldn't have said 'meaning'. more straightforward to just
> say the ternary relation has a (implied) constraint that the two binary
> ones don't.

Ok (in 'no nulls' mode).

> actually several - eg if employee has a number, then
> employee must have a name and an address.

How so?

*If* you would say a row in E (you snipped the part where you defined E - hard to discuss this way) means there exists an employee with that number, (and, assuming a closed world, that there is no other way to state that fact) - yes.
But you said no such thing (yet?).

>>> 2) if i natural join N and A, producing a result that matches the 
>>> contents of E, am i introducing the constraint in some sense?
>>
>> Only for the resulting view - but I'm not sure what you mean here.
>> If it is enforced as a constraint - how could you
>> insert into A - or into N?

>
> i guess if the constraint, implied or not, applies only to the view then
> inserts to A or N wouldn't be affected. if it were somehow enforced for
> A or N (for example, if A.emp# was a foreign key E.emp# and vice-versa,
> even if that seems a bizarre way to do it and similarly for N.emp#),
> then some 'transaction' mechanism would be needed for separate insert
> 'statements' to succeed, but that's another topic, i think.

Are you indicating a "'transaction' mechanism" where all constraint checking is deferred until the end of the transaction?

> perhaps the implied constraint i'm talking about is nothing more than a
> fancy way to say that nulls aren't allowed.

Perhaps. Could it be a consequence of not allowing nulls? Received on Sun Jul 10 2005 - 03:47:50 CDT