Re: NULLs: theoretical problems?

"Jan Hidders" <hidders_at_gmail.com> wrote in message news:1187305514.703264.214690_at_d55g2000hsg.googlegroups.com...
> On 17 aug, 00:03, Hugo Kornelis <h..._at_perFact.REMOVETHIS.info.INVALID>
> wrote:
>>
>> [....big snip ...]
>>
>> And the model is extremely unstable, since it's not uncommon in real
>> businesses to see an optional proposition become mandatory (e.g. ebcause
>> of legislation change) or a mandatory proposition become optional (e.g.
>> because of competition). If all propositions are combined into a single
>> table, this is a simple change from NULL to NOT NULL or vice versa. If
>> each proposition has its own table, this is an equally simple change,
>> the addition or removal of a constraint that SQL does not currently
>> support but that is known as "equality constraint" in Object Role
>> Modeling.
>
> Just a minor technicality. As far as I understand it an equality
> constraint can be represented by two inclusion dependencies in both
> directions, so I would say that it *is* supported by SQL. Anything I'm
> missing?
>

Yes. SQL doesn't support multiple assignment. In order to get rows with new key values inserted, at least one of the constraints needs to be disabled.

> Btw. your interpretation of null values is basically the "missing
> value" interpretation, with which theoretically speaking, there is of
> course no problem whatsoever. But your insistence that NULL somehow at
> the same time a marker that indicates that a value is missing and is
> also a value itself (because an element of the domain) is a little bit
> suspect. There are better and more consistent ways of describing
> missing values.
>
> -- Jan Hidders
>
Received on Fri Aug 17 2007 - 14:17:21 CEST