In assembly language, a sub-procedure can save the state of all the registers it will use before using them and then set the state back when it is done (push register values on the stack, do work using those registers, and then pop the state back). This is the safe thing to do, but it wasn't the best way to write code back then. People would use unsafe procedures instead to reduce code size. With that said, I think Bill was talking about the fact that if the program doesn't repeat itself (by using procedures to abstract common idioms), then a mistake would be amplified since it would cause problems for all call sites.