[SystemSafety] State of the art for "safe Linux"

[Derek M Jones]

Paul,

> You may be right - certainly so far I haven't found much in the way of published research to counter your statement.

I claim that my book "Evidence-based Software Engineering"
discusses all the publicly available data, as of 2020 (not that
I have found much since),
pdf+code+all data freely available here:
http://knosof.co.uk/ESEUR/
which sounds ridiculous, but there is just not that much
public data.
See the reliability chapter, the one for which I found the
least amount of data.

> And interesting for me in that it clearly supports the argument that (the chosen version of) Linux's timing was non- 
> deterministic.

The Experiments chapter of my book discusses examples of performance
variability of the same program across reboots of Linux (it's larger
than rerunning within the same booted system).

>> Allende's analysis makes various assumption that the available data
>> suggests don't apply to software reliability.  I'm happy to talk
>> about this in another thread.
> 
> Yes please!

People assume that software fault experiences follow a Poisson
distribution, which they often do for hardware.  The data suggests
that software fault experiences have an exponential or power law
distribution.  Some analysis
https://shape-of-code.com/2018/03/25/top-must-read-paper-on-software-fault-analysis/
https://shape-of-code.com/2017/12/12/the-shadow-of-the-input-distribution/

These different distributions probably have a big impact on
the analysis and the results.  I have not seen any analysis using these
distributions (not sure that my maths pay level is up to doing it).

>> My PhD thesis work is ground breaking.
> 
> Sorry, I'm being dumb here. Do you mean that $PhD student always concludes that their work is groundbreaking?

These days, yes.  In fact over-the-top claims of the importance
of one's work are now an integral part of academic publication.
Reading papers from the 60s/70s is a real culture shock, they
are so genteel.

>> My question is how this Linux variability compares with the variability
>> that must also occur in other operating systems?
> 
> That's a good question for sure... probably at least a couple of PhDs in that :)

Researchers use Linux because it's the done thing, and they think
that open source will make their life easier.


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