Interesting read: [WayBack] Software quality: economics of formality
At one time formal methods were thought to be the only hope for software development. That did not turn out to be the case.
Via: [WayBack] Software quality is better in practice than in theory. Formal methods improve software quality, but at a high cost that is sometimes worthwhile and some… – Jeroen Wiert Pluimers – Google+
Comment there:
+Jeroen Wiert Pluimers +John Cook , I think that when a code has been in service for even a short time, we can divide the code into two classes. The first is that part that has been executed at least once. The other code has never been executed. The latter code has an error rate that may be measured as a function of the language used. In the old days Fortean had a bug in every seven lines. C used to be 100 to 200 lines. We set up instrumentation to determine the number of lines tested in the first class and found that it was a linear function of the new bugs found. This was a good way to ensure a high quality code.
Related:
- [WayBack] The Shape of Code » Mathematical proofs contain faults, just like software
- [WayBack] The Shape of Code » formal methods
–jeroen