Well, your first and third question are basic database design issues not related to Oracle specifically. Personally, I can't imagine the reason for having tables without PKs. I make an abstracted, or surrogate, key for (nearly) every table in every database I make. These are simple, 1 column, numeric keys filled via sequences. This eliminates the question of uniqueness. However, I am concerned that you ask about that at all. If you would not have a unique PK, how would you ever identify an individual record? You would not even be in first normal form. As for the FKs every FK should be defined, except under extreme circumstances. Constraints and indexes can be disabled or dropped at imp/exp time if that is an issue, but sound database design would require their existance.

Of course, I am assuming an OLTP database. If you are doing DSS, then modifications are not an issue, so the FKs drops in importance. However, you still have to join tables, so if you have a child table, it still has to have the PK from its parent, so you still have to have PKs.

Sorry I don't have anything for question 2.

HTH,

- Chris