Re: NULLs: theoretical problems?

"David Cressey" <cressey73_at_verizon.net> wrote in message news:lxyxi.34$wr3.33_at_trndny04...
>
> "David Portas" <REMOVE_BEFORE_REPLYING_dportas_at_acm.org> wrote in message
> news:l_Cdnbfs5KQfllvbRVnytgA_at_giganews.com...
>> "David Cressey" <cressey73_at_verizon.net> wrote in message
>> news:UGmxi.11$dz3.4_at_trndny01...
>
>
>> SELECT DISTINCT i FROM t WHERE x = x;
>>
>> The result is clearly wrong if a null means "x is unknown" or "x is out
>> of
>> range" and possibly wrong if null means "x does not apply".
>
> But if a null means "x is not present" and nothing more, the result is
> clearly right.
>

Giving yet another interpretation of null demonstrates only that there is no generally accepted interpretation. What on earth is "not present" supposed to mean in a proposition anyway? I suggest that it is far from clear that "not present" means that x = x evaluates to something other than true!

>>
>> If you admit nulls then you must deal with predicates that reference
>> nullable attributes.
>
> Why?
>

You are seriously suggesting that nulls should never appear in predicates? That alone must be a strong argument to banish them entirely.

-- 
David Portas