Michael Glaesemann wrote:
>>> In any case, one would never use NULL. Either the domain includes a
>>> value for all possible values (including N/A) or you set up the db
>>> schema appropriately.
>>
>> Hm, that can be painful. What if I have ten optional attributes;
>> separate them to ten different tables?
>
> Strictly? That's how I understand it. Whether or not it's performant
> under PostgreSQLis another matter. :)
The relationists' SQL replacement had better have more convenient JOIN
syntax then. :-)
--
dave