Hey all - can I firstly just join the "I think this is a good article" camp - obviously there are always more than one way of doing things - but assuming you're not always using 2005+ I think this solution is very nice.

One thing I stumbled on in the following discussion this comment "It's not always required to add unique indexes/constraints, though that was a good tip."

Just out of curiosity - if I have a table where I use surrogate PK of some sort (Ints or GUIDs or whatever) I always make sure that I also have a natural key in the form of a unique index - so say it's a Person table I might place this on the email, if it's an Order table I might place it on the Customer and TimePlaces etc.

I normally go to quite some length to do this, mainly just because I was once taught this was good practise - and also because I find it traps a lot of application logic errors that would otherwise go unnoticed.

Now I should say that I am an application developer and not a DBA - so I am actually quite interested in hearing your opinion on this.