When I did software testing for one of the UK high street banks we had a
theory about software bugs. We reckoned that the rate of fixing bugs in
a complex piece of software was approximately linear, but the rate of
finding new bugs decreases because there should be fewer bugs to find
(as long as the attempted bug fixes didn't introduce more new bugs as
side effects). By observing the way the number of known outstanding bugs
decreased over time, we considered it possible to estimate the probable
number of unknown bugs remaining in the system.

So here's the numbers of outstanding confirmed bugs from the various
known bug lists.

        Date            known bugs
        ----            ----------
        27 Sep          19
        28 Sep          16
         1 Oct          21
        23 Oct          25
         1 Nov          27
        21 Nov          38
        27 Nov          42
         2 Dec          37
        current         42      (list not yet published)

We appear to be moving in the wrong direction.
(That never happened at the bank.)

-- 
Mike Williams
Gentleman of Leisure

Post a reply to this message