Investment in reliability will increase until it exceeds the probable cost of errors -- or someone insists on getting some useful work done.
All programs contain errors until proved otherwise, which is impossible.
Undetectable errors have infinite variety; detectable errors by definition are limited.
If nobody uses it, there's a reason.
In a bureaucratic system, useless work drives out useful work.
Merely because the group is in formation does not mean the group is on the right course.
Things should be as simple as possible -- but no simpler.
Once you have eliminated the impossible, whatever remains, however improbable, must be the truth.
What, me worry?
The effectiveness of a telephone conversation is in inverse proportion to the time spent on it.