Sometimes, IDENTITY columns are a Bad Thing for uniqueness (AKA, why I like uuid instead). Why? Well, three reasons IMHO.
- They're finite. Let's take a shopping cart as a simple example. If I have an IDENTIY/AUTOINC column for each item added, that adds up over time. Will I max out that BIGINT field? Eventually. Granted it may be a loooong time down the road, but why build in that limitation to begin with?
Uniqueness. :-) What's the difference between row 243 in one table and row 243 in another table? Sometimes, nothing except their uuid. For example: db data change logs in an app. If each row has a unique id, then I can use it as a foreign key in the log data table. Sure I could use numbers, but uuids provide a little more assurence that the foreign key IS correct, instead of coincidentally matching by number.
Sometimes, IDENTITY/AUTOINC columns are pure evil. I've seen on more than one occasion that the ident/autoinc seed for a table gets reset, after which time the next IDENTITY number may actually be from a record that was previously deleted. Next thing you know a join is fubar just because a new entry aquired an old identity seed. It shouldn't happen, but it does sometimes.
My point? Good question. :-P