>>>> Attempts to mask NULLs with unknowns or any other value,
>>>> don't work. (you must not have read my prior posts
>>>> expounding the above point).
>>>
>>> Let's say ... value 0 to indicate unknown age.
>>
>> If two things have age 0,
>> is it true or false that their ages are equal?
>
> If we allow for unknown values, then we necessarily must also allow
> for functions invoked on those unknown values to returns unknown
> values. On the other hand, if we treat missing information as explicitly
> not being there, (a situation that I find slightly preferable) then
> the sum of an optional age, not present, with another age, not
> present, results in an optional integer, not present. Cardinality
> then becomes a different "dimension" as it were to types/values,
> and we stay 2VL.
Please read, Missing Information in CJ Date's "Intro to Db Systems"
(Chapter 20 in his 6th Ed). Maybe you'll understand why the last line
of the chapter states "NULLs and 3VL undermine the entire foundation
of the relational model" (pg 587).
Received on Thu Nov 18 2004 - 23:46:28 CST
