[SystemSafety] Difference between software reliability and astrology

[Prof. Dr. Peter Bernard Ladkin]

On 2024-08-13 12:28 , Derek M Jones wrote:
> Martyn,
>
>> A rational model to connect the data with the conclusions
>
> What counts is models that work.  We rationalize them
> after it has been shown that they do a good job of predicting
> our reality.

That's not the way it works with many or even most models. In particular, that is not the way it 
works with the reliability models used for safety-related software for which there are what 
Littlewood and Strigini called ultra-high dependability requirements.

For on-demand functions, various characteristics of the use of the software lead you by rational 
argumentation to a Bernoulli process (the Bernoulli process being the only discrete process which 
satisfies the selected characteristics). Similar considerations for continuously-operating software 
to a Poisson process (or to some renewal process for components, whereby by a limit theorem the 
entirety behaves Poisson-ly). There is no matching of results ("doing a good job of predicting .. 
reality) because the outcome you are looking for is an absence of something, namely failures. If you 
see one, or more, it's back to the drawing board with the software.

It is, as Martyn said, rational. Ratiocination. Inference. Whatever one's favorite word for it is. 
It is not rationalisation.

PBL

Prof. Dr. Peter Bernard Ladkin
Causalis Limited/Causalis IngenieurGmbH, Bielefeld, Germany
Tel: +49 (0)521 3 29 31 00