[SystemSafety] "FAA chief '100% confident' of 737 MAX safety as flights to resume"

[Olwen Morgan]

On 30/11/2020 00:59, Les Chambers wrote:

<snip>

> In the abstract, should we trust software to make any unstable real-
> world system of any kind, stable, predictable and safe?
> As a card-carrying stoic I offer that the no voters are about to be overrun by the reality of apps
> such as driverless cars where software failure will , with high probability , result in injury and
> death.
> Further, be advised that the yes voters have been operating for some time.

<snip>

YES voters bear the obligation of ensuring that the interfaces between 
the various other systems engineering disciplines and software 
engineering are well controlled; that is, on each side of a critical 
interface, both kinds of engineers have an unambiguous understanding of 
what the software is to achieve and what stress conditions it must 
withstand. If that degree of control is achieved, then, *on  a 
case-by-case basis*, I would not *necessarily* oppose the use of 
software to compensate for the properties of a simple but not entirely 
robust physical design.

But the NO voter in me simply observes that the degree of technical 
control that this implies is one that has been very rarely achieved in 
my working experience of critical systems engineering. In practice, 
therefore, a software engineer who objects to clearing up the ordure 
left by other disciplines is simply exercising an eminently reasonable 
right not to be landed in the said ordure where they may be blamed for 
things that were not their fault.

In general, I have found interface control to be the weakest aspect of 
software development processes with which I have had to work.


olwen