The error was traced to also exist in previous version of the software, but had caused no problems because there was only one mode in that version. Was it the designer or the tester (who looks for errors and unconformities to the requirements) who was at fault? The Therac case also illustrates the difficulty of testing real-time systems. Should there have been a feature on the system limiting the radiation dose to some MAXIMUM irrespective of everything else. If so, should this not have been specified by the clients? Are they also partially to blame? soulution