Maintenance needs 50% life-cycle cost, modifications and fault removal. Once the fault is removed, how do we test the system after competing the maintenance? You know, every maintenance leads some are introductions, some of them are faults. So we should be able to test it to see that the system is dependable at after modifications. So, should we have total retesting? Or selective retesting. And who will do that? Do you expect the maintenance test has to be as good as developers do the testing? Do they understand the total system? So, those are the issues.

---