[SystemSafety] The dilemma of software process and dependability measurement

[Olwen Morgan]

On 16/11/2018 15:13, Steve Tockey wrote:
> Of course I wish I could have built several of exactly the same system in parallel. Unfortunately I have only been able to do that once. You realize it is too expensive to do that in the general case.


Actually, it's worse than Steve T suggests here.

If you compare the results from two processes, one with state-of-the-art 
maturity and one of lesser maturity, then you immediately hit a 
methodological dilemma. If the less mature process is so uncontrolled 
that it is not reliably measurable, then you lack the necessary 
metrological basis for a controlled comparison of process capabilities. 
On the other hand, if the less mature process is at least mature enough 
to permit reproducible measurement, then observable differences in 
results may not be particularly salient.

That's why I never try to justify best practice techniques on the basis 
of their contribution to process capability. I only ever try to justify 
them on grounds of cost-saving due to earlier detection and correction 
of errors.


Olwen