[SystemSafety] Two Preprints

[Derek M Jones]

Peter,

> Evaluating Software Execution as a Bernoulli Process DOI: 10.13140/RG.2.2.28079.82082
> 
> This is my response to the odd view of Daniels and Tudor in their recent SCSCeJ paper that evaluating discrete-function 
> software execution as a Bernoulli Process does not work (while acknowledging that maybe it doesn't work for them :-) ).

Let's say that a software system contains M coding mistakes.
Each is likely to have a different probability of being
triggered by some input, causing the system to fail.

If we treat the probability of each mistake as having
a Bernoulli distribution (I think that exponential
is a better map to reality, but let's put that
to one side), then by Le Cam's theorem, the system
failure distribution has a Poisson binomial distribution
https://en.wikipedia.org/wiki/Le_Cam%27s_theorem

A Bernoulli Process is great for modeling coin tosses and
very very simple systems.

Daniels and Tudor discuss realistic systems and reality
is more complicated than tossing a single coin.

-- 
Derek M. Jones           Evidence-based software engineering
blog:https://shape-of-code.com