Re: circular relationships ok?

>> This is an implementation detail that can be got right or got wrong, not an argument for refusing to support the rules of the business. <<

The problem is logical, not implementation. In order to get a deterministic result, you would need to deter constraints in the same order, every time. I suppose the right answer for my A-B-C example would be to detect the cycle and do a ROLLBACK when they conflict, but run the changes in other cases.

>> (It is also, dare I say, a glaring example of why in the 21st century we should be asking why we are still satisfied only to be able to make one incremental update to one row of one table at time.) <<

Unh? We can update a **subset** of a single base table at a time; that is not a row at time. We did get BEGIN ATOMIC ... END; for transactions and defered constraints with SQL-92, so you can do a lot of things.
. Received on Thu Mar 02 2006 - 17:46:29 CST