[SystemSafety] New paper on MISRA C

[Derek M Jones]

Mallory,

> What would be the practical utility of indefinite probability data about the correlation between safety and reliability*, and is that function best served by such data?

Probability distributions are needed to build models and
help answer some kinds of what-if questions.

I have not had model building or what-if question related
experiences with safety/reliability.  Regular readers will
know that I collect data (you can never tell when something
is needed), the question of correlation came up and so I
asked about data.

> The kind of indefinite probability data we’re talking about would support the claim that a system need not be reliable to be safe and vice-versa, of course.  But such support seems superfluous given the examples already in hand.
> 
> Indefinite probabilities are sometimes used as estimates for definite likelihoods of interest** such as the likelihood that a failure in a given system would compromise safety.  But wouldn’t a causal analysis of the system in question reveal much more about whether and how failure of a given subsystem/component/service/etc. might impact the system’s safety than information about peer systems?
> 
> Kind regards,
> — Mallory
> 
> Dr. Mallory S. Graydon
> Research Computer Scientist
> NASA Langley Research Center
> 
> 
> *  Presuming, of course, that it is possible meaningfully measure the safety of rare, high consequence events.
> 
> **  As an example of where this can go wrong, an insurer declined to cover my home on the grounds that my fireplace insert is more than 20 years old.  No doubt they’ve got indefinite probability data showing a higher-than-acceptable correlation between high fireplace-insert age and home loss through fire.  But what they don’t know is that I never use my fireplace.  Thus the definite likelihood that I will experience a fire due to failure of my fireplace insert is exactly zero, regardless of other insured’s experiences.
> 

-- 
Derek M. Jones           Software analysis
tel: +44 (0)1252 520667  blog:shape-of-code.coding-guidelines.com