"Keith F. Lynch" wrote:
[...]
Formal research by experts in N-version programming have found
the same thing: N-version programs are *not* N-times better than
1-version programs.

I'm not sure what N-times better *means*. What twice as good as 99%
reliable? 198% reliable?

Turn it around -- 99% reliable means 1 failure in 100 tries. Twice as
good would be 1 failure in 200 tries (or 1/2 in 100) or 99.5%
reliable.

/dps