Posted by Bill Karwin on 03/07/06 19:51

"Ed Prochak" <edprochak@gmail.com> wrote in message
news:1141752316.933161.102260@i39g2000cwa.googlegroups.com...
> My point is too many people immediately jump to ID fields as the PK.

That's certainly a good point. However, I've worked on projects in which
the decision-makers wouldn't commit to _any_ combination of attributes that
would uniquely identify the entity. There were always cases where the value
in any column could be either non-unique, or else have no value specified
(i.e. NULL). Neither would they commit to any attributes that could be
reasonably stable and unchanging (though I understand that this is not
strictly necessary for a key).

So in those kinds of situations, I felt I had to create pseudokeys to have
any chance of the application working. Even if we know the best practices
for database modeling, the project on which we are working may have
constraints that don't allow us to follow those best practices.

Regards,
Bill K.

[Back to original message]