Re: Unique constraint and NULL values

Frank Piron <empty_at_zero.nil> wrote:

>Thu, 21 Oct 2004 11:40:48 -0400 Mark C. Stock <mcstockX_at_Xenquery .com>
>wrote:
>>
>> yes, in theory, the SQL Server approach is more 'pure', but in practice,
>> the
>> Oracle approach is more practical (i'm starting to sound like yogi
>> berra!)
>
>Agreed. IMO the most plausible argument for the Oracle approach
>is the observation that allowing exactly one row with a null value
>in a certain column would mean that
>1)
>The column value is "identifying" the row (unique constraint).
>2)
>NULL - an undefined value - is an "identifying" value.
>
>which does not make sense for me.

It can get almost metaphysical, but:
Identifying a column which has an undeterminable value ( NULL) is not the same as identifying that value, so being able to consider that column unique ( that is , it is the only one with the undeterminable value) is not a problem. Received on Fri Oct 22 2004 - 15:33:52 CEST