Each unit test must be run independently of other unit tests. Otherwise it will not be portable, it is difficult to maintain, not clear enough. Errors must be monitored and reported in units of single items. It is easy to define which unit tests to run.
Excerpted from "junit in action" P 10
