> This may be true for some use cases, but in general that's wishful thinking.
Why would avoiding implementation of database level constraints be "wishful thinking"?
Au contraire: I strongly believe that data level constraints should be tied to the database.
Sure, there's a trade off between indexing, constraints and performance. But I rather have an additional unique index on a column than relying on the fact that all developers in an organization always do the right thing.
If the database is my responsibility then I'll make damn sure that it's not possible to fuck it up with shlocky code.
Why would avoiding implementation of database level constraints be "wishful thinking"?
Au contraire: I strongly believe that data level constraints should be tied to the database.
Sure, there's a trade off between indexing, constraints and performance. But I rather have an additional unique index on a column than relying on the fact that all developers in an organization always do the right thing.
If the database is my responsibility then I'll make damn sure that it's not possible to fuck it up with shlocky code.