CMOS reliability: a useful case history to revise extrapolation effectiveness, lenght and slope of the learning curve