john2, that's a good one. No it's I'm not looking for a end all be all, but it is a nice feature, especially for bulk insert operations when you repeat then.
How many times have you tried to load data that had a duplicate PK and it fails. Be nice to have the upsert work on a set of data based on the PKs. Like with an update, you would supply the PK and other values.
So in the Northwind db.
values (employeeid, lastname)
select 8, 'Jones'
This would check the employees table for an employeeid (PK) of 8. If one exists, the last name is updated to 'Jones' and everything else is left alone. If it does not exist, a new row with employeeid = 8 and lastname = 'Jones' is inserted. This would be subject to the same insert rules and would require defaults or nulls allowed for all other fields.