Date: 29 Dec 98 03:12:44 From: firstname.lastname@example.org (James Matthew Weber) References: 1 2 3 4 5 6 7 8
View raw article or MIME structure
>Interestingly, one of the SW QC efforts that I heard about years ago >(sorry can't remember the source but I believe it was in a GD proposal >to the Army DIVAD program citing their QC efforts on the F-16 program >and what they had learned via AI) was that AI had three separate >development organizations develop the same functional code. These >three organizations did not know the others existed and were located >in three different countries. The theory behind this was that AI >would catch more bugs in critical flight control sw this way because >each group would develop differently, discover different bugs, and >would provide AI with better de-bugging capabilities (hence higher >quality sw) when the time to integrate the sw into the hw took place. >Oddly, what they found was that bugs detected during testing were very >much the same across groups with few exceptions, even though the code >may have been significantly different. This really shouldn't come as any surprise. The people who write software have often undergone very similar training, and tend to solve problems the same way. The result is that all solutions are likely to be built on, and contain some of the same, underlying assumptions. There is no assurance that thos assumptions are good or accurate. Earlier in my career I had a situation where two organizations had to solve the same problem. They didn't like each to the point that I'd characterize them as not being on speaking terms. When you looked at the Assmebler code produced, you would have sworn the same people wrote both. the methodology and solution were virtually identical.