gutfreund.umass-coins@UDel-Relay@sri-unix (11/22/82)
From: Steven Gutfreund <gutfreund.umass-coins@UDel-Relay> Not only did they have Rockwell write the backup software for IBM, but they paid Rockwell a bonus for each bug they found in the IBM software. I thought this was a very cute way to try and overcome the "software reliability" crisis, -> pay a competitor to find the bugs in your programs. - Steven Gutfreund (roi de soleil)
REM@MIT-MC@sri-unix (11/23/82)
From: Robert Elton Maas <REM at MIT-MC> I think that's a great idea! I wonder if they also paid IBM for finding bugs in Rockwell's software?
dave (11/24/82)
Paying a competitor to find your bugs is indeed a good way to encourage the IBM guys to write good software and to reassure yourself that the software is fairly correct. But it worries me a bit that the people writing the backup software were even looking at the primary code. One approach (which apparently wasn't used) to developing fault-tolerant software is called N-version programming [Chen and Avizienis]. N different versions of the same program are written independently from the same specifications; during execution, all N version are executed, the results are compared, and the majority answer accepted. One of the safeguards in having someone else write the backup code is that it should be done independently, without bias. Thus, if IBM made some subtle fault in the structure of their program, the fault might also have been inherited by the Rockwell software. With their budget constraints, NASA appears to have been more concerned with the primary software--getting that right--and hoping that they never have to rely on the backup software.