In article <3E2808A5.29237234_at_nll.ibmm.com>,
Anton Versteeg <av_at_nll.ibmm.com> wrote:
>Jan Hidders wrote:
>>
>> The only requirement for a key is that *at every moment* there are no two
>> entities (tuples) for which these values are the same. Since the firm
>> name will be unique before the change and after the change it is
>> therefore a good key.
>
>Sorry, I disagree: names are IMO not good for primary keys. Often they are
>not unique and it is easy to enter wrong names (typo's etc) and thus wrong
>keys.
Well, if they are not unique then they are obviously not a key by
definition. :-) The point was that if they are unique then you could use
them as primary key. The fact that names are not always typed right is a
problem that does not go away if you introduce surrogate identifiers.
Received on Fri Jan 17 2003 - 08:46:32 CST
