>Then, clearly, if a small change to the production code causes large changes to the tests, there is a design problem. > >Therefore the tests must have their ow

>Then, clearly, if a small change to the production code causes large changes to the tests, there is a design problem. > >Therefore the tests must have their own design. They cannot simply follow along with the structure of the production code. The jump indicated by the "therefore" confuses me. I essentially couple my test to production behavior, insofar as it is possible. I do this to test program behavior realistically and because I believe tests should, in some sense, form a specification for the program they test. I don't really care about "testing code"; I care about testing behavior. I don't have the problem of small changes to code necessitating large changes to tests. It makes me wonder what he does that is causing that.