JOG wrote:
> Bob Badour wrote:
>
>>Marshall wrote:
>>
>>>On Jan 8, 4:44 am, "David" <davi..._at_iinet.net.au> wrote:
>>>
>>>>My point is that the following six conditions can't all be satisfied at
>>>>once
>>>> C1. use person(P,M,F) relation
>>>> C2. don't allow nulls in M,F
>>>> C3. enforce referential integrity on M,F
>>>> C4. only allow finite number of persons in the domain
>>>> C5. there are no cycles in the family tree
>>>> C6. there is at least one person
>>>>
>>>>Obviously we must have C4, C5 and allow C6. I suggest that C2,C3 are
>>>>important and therefore C1 should be dropped. ie the person(P,M,F)
>>>>relation itself is "bad". Do you agree?
>>>
>>>This analysis looks right to me.
>>
>>Huh? Your solution basically achieves all of the above with the
>>inconsequential difference that you used a different name for the
>>relation in C1.
>>
>>Nowhere above does it mention the very consequential difference in
>>referential integrity constraints, which is in fact the straw in the man.
>
> Hi Bob, if you were referring to my post then your last statement isn't
> true. I emphasized the change in referential integrity constraints as
> being important:
You emphasized it and then David ignored it in his reply. If you look
carefully, I think you will agree what you wrote appears "nowhere above"
where I wrote "nowhere above".
> JOG wrote
>
>>where P is the candidate key of _parentage_, but where P, M and F have
>>enforced referential integrity (with a check constraint on sex) back to
>>the _person_ relation
>
> Either I've missed your point or you had a heavy new years celebration
> this year!
You missed the point that David constructed a straw man by ignoring your
point entirely.
Received on Tue Jan 09 2007 - 07:27:22 CST
