[SystemSafety] Putting Agile into a longer perspective

[Olwen Morgan]

On 21/10/2019 20:41, Derek M Jones wrote:
>
> Lots of people have proposed theories, it's a shame that none are based
> on evidence (not that there is much evidence around).
>
The difficulty of obtaining hard reproducible evidence of the 
effectiveness of development practices is a perennial curse of software 
engineering. On the other hand, there are cases where what theory 
predicts should work actually turns out to do just that when it is 
tried. The accomplishments of SPARK Ada are a case in point.

While, if we are trying to be strict scientists, we should prefer 
evidence-based evaluations over all others, we find ourselves at a stage 
where we have to try things that are theoretically reasonable simply to 
generate the evidence we wish we had at the start.

On occasion, I find myself thinking Derek an insufferable Jonah when he 
bangs on about evidence. That's not because he isn't on rock-solid 
ground but simply because the circumstances in which software 
engineering is done are simply not able - at least yet - to yield 
robustly reproducible data. We are not, I think, even at the stage when 
competent meta-analysis of diverse studies will actually give the 
robustness of evidence we would really like. (Sadly, in this respect, 
software engineering resembles the dismal science of economics.)

Yes we should always have in mind the need for evidence - but this 
should not prevent us from trying out things that look promising and 
then, if necessary, suspending judgement for long enough for clear 
evidence to emerge. Nor, should we, get too hung up on whether evidence 
is robustly reproducible in circumstances where confounding factors 
virtually preclude that possibility. We have to improve software 
engineering starting from where we are. If evidence is currently 
lacking, we simply have to content ourselves with the long haul of 
amassing it painfully slowly in adverse circumstances.


Olwen