[SystemSafety] IEC 61508 Committee Draft of Edition 3 published

Derek M Jones derek at knosof.co.uk
Tue Oct 4 20:28:06 CEST 2022


Peter,

> Bertrand Ricque and Michael Kindermann (both on this list) are both thinking about an IEC project to write a standards 
 > document with more detail on the statistical evaluation of safety-critical software.

This sounds like a good idea until it is realised that
we know practically nothing about the reliability of
software.

One major problem is the almost complete lack of
empirical software testing research.  Yes, there
are lots of researchers publishing arm waving mathematical
papers, but the empirical side of things is a lot
less popular.  Where is the research extending Nagel and
Skrivan's fantastic work?
https://shape-of-code.com/2018/03/25/top-must-read-paper-on-software-fault-analysis/

Governments fund labs to stress test physical objects.
I'm not aware of any such funding for software, and there
ought to be.

> document with more detail on the statistical evaluation of safety-critical software. There seems pretty much universal 
> agreement that Part 7 Annex D is not enough. However, there is a lot of disagreement about what more can/should be said. 

These four cases caught my eye.  I can see the logic behind
D.2.1.1 and D.2.3.1, i.e., test components in proportion to
likely usage.

Is the purpose of D.2.2.1 to keep costs down?

What happened with D.2.4.1?  Did a bunch of managers turn up
for a drive by shooting when this got written?

D.2.1.1
a)  Test data distribution equal to distribution for demands during
on-line operation.

D.2.2.1
The only prerequisite is that the test data is selected to give a random
uniform distribution over the input space (domain).

D.2.3.1
a)  Test data distribution equal to distribution during on-line operation.

D.2.4.1
a) Test data distribution is such that each of the N program properties
is tested with equal probability.

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


More information about the systemsafety mailing list