[SystemSafety] Bossavit's Leprechauns book

[Olwen Morgan]

On 07/12/2018 07:47, Nick Tudor wrote:
> <snip>
> The experiment was to measure the time taken to satisfy the review 
> process between requirements and design in Simulink.
>
> Not only did our tool find all the errors, we also found an extra one. 
> Against the benchmark savings of 60-80% were found and against the 
> incumbent 50-60%.

<snip>

IMO the best arguments in favour of using tool-supported formal methods 
are those based on the efficiency with which they detect errors. The 
results you quote for a controlled trial do not surprise me.

There is, IMO, no good reason to espouse weak arguments that such 
methods lead to better dependability properties if they can be 
adequately justified by what, essentially, are savings in quality 
control costs.

Of course, deniers may object that the circumstances of the experiment 
do not reflect real conditions of use but at the very least your study 
points the way to how one may set up rigorous trials in the normal 
working environment.

Olwen