As system architects we have to foresee things that can change in the software specification over time. We have learned that a good system design is one that seperates the things that change from the things that do not. However in trying to foresee potential changes we often try to design for changes that may not happen. This leads to an over-engineered design which is more time consuming to implement and difficult to understand. Such a design becomes counter productive. While designing a software it is important to view the changes we anticipate along with their probablity of occurence. It is often a good idea not to design for a low probability change, if it significantly increases the design complexity.