Walter,

>I allowed changing the intelligent candidate key at first, but then disabled it in most cases, because some users did thoughtless changes, without being aware of the implications: since they couldn't delete a record (RI!), they overwrote it... argggghhhh

Yep, this is a real pitfall and one that should be taken care of when dealing with surrogates. You've dealt with this situation, I wonder how many applications are waiting to fall into the same pitfall.

>Rushmore Optimization is really very simple: "Use every available index, whether it makes the query faster or not" - since they found it difficult to decide a priory whether an index indeed speeds things up. A database server with rule-base or, better, cost-based optimization is more efficient here, AFAIK.

Cost base optimization is often difficult to implement. The database system has to maintain some statistics about key distribution etc.

>I wrote something about this in FAQ#8109.

As I recall correctly I did participate in a thread leading to this faq. Yes I agree with your conclusions. However about 3 years ago, only few of us knew this.

>Met hartelijke groeten, Hilmar.

I knew, you've got dutch blood. The name zonneveld is too dutch to be bolivian. How did you get there ?

Met vriendelijke groet,

Walter,