[SystemSafety] Recent over-the-air update improved the car’s braking
Bev.Littlewood.1 at city.ac.uk
Thu May 31 17:24:26 CEST 2018
I have some sympathy with Derek’s view, but I’d put it a bit more strongly. Rather than his “warm feeling” from some testing, I'd prefer there to be “sufficient” testing - i.e. to be able to make a claim about its reliability/safety with a high enough confidence. Which I think is your implicit point: how much is “sufficient”?
Without going into the numbers (which always seems to be like poking a hornets’ nest on this forum) I think we could all agree that in the time that elapsed only very weak claims from testing could be made (if indeed they did any testing at all).
I suppose they might say that they “just” tweaked some parameters, with the implication that there were not substantive changes to the software. So that the results from any extensive software testing they had done in the past (if indeed they did this) could still be relied upon. But this won’t wash, will it? We would need to know how extensive the parameterisation is here, and whether all possible parameterisations had been tested to support the required reliability/safety claims.
Many years ago I was involved in the controversy surrounding the safety claims for the software-based Primary Protection System of the Sizewell B nuclear reactor here in the UK. The “software” was a couple of hundred thousand lines of code. But the “configuration data” was of comparable size. Testing the system in just one configuration was problematic. Gaining confidence in its safety in all configurations was very hard.
I wonder whether this is an issue in the Tesla case? Seems to me it might be.
Of course, an important difference in these situations is that in the Sizewell case a single nasty event needs to be highly improbable. In the Tesla case, maybe they (we?) can live with a couple of nasty events. After all, if this happens, they’ll fix the problem, won’t they? I.e. they’ll tweak it after a few incidents (maybe deaths). But we would have the same sort of doubts about the efficacy of this tweak as we have had in the current one. And so on...
On 31 May 2018, at 09:07, Martyn Thomas <martyn at thomas-associates.co.uk<mailto:martyn at thomas-associates.co.uk>> wrote:
What specific objectives should such testing have? How much and what
sort of testing would deliver the required evidence that the objectives
had been achieved?
On 31/05/2018 01:35, Derek M Jones wrote:
No mention of testing:
The System Safety Mailing List
systemsafety at TechFak.Uni-Bielefeld.DE<mailto:systemsafety at TechFak.Uni-Bielefeld.DE>
Professor of Software Engineering
Centre for Software Reliability
City, University of London
Phone: +44 (0)20 7040 8420 Fax: +44 (0)20 7040 8585
Email: b.littlewood at csr.city.ac.uk<mailto:b.littlewood at csr.city.ac.uk>
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the systemsafety