On 2026-03-05, Carlos E.R. wrote: > On 2026-03-05 18:42, Charlie Gibbs wrote: >> On 2026-03-05, Peter Flass <[email protected]> wrote: >> >>> On 3/4/26 15:35, Lawrence D’Oliveiro wrote: >>> >>>> On Wed, 4 Mar 2026 14:09:58 -0700, Peter Flass wrote: >>>> >>>>> On 3/4/26 13:29, Carlos E.R. wrote: >>>>>> >>>>>> On 2026-03-04 21:01, Ted Nolan <tednolan> wrote: >>>>>> . >>>>>>> A man with one clock knows what time it is. A man with two is >>>>>>> never quite sure... >>>>>> >>>>>> Experimental science would not agree. >>>>> >>>>> You would need at least three. >>>> >>>> Three would be better than two, but two is already enough to come up >>>> with an error estimate on the measurement. >>> >>> No, because if two don't agree, one could be just plain wrong. The Space >>> Shuttle system had three processors run the same computation as a check. >> >> And, IIRC, the third one was built and programmed by a different outfit. > > Different software? That one I did not know.
Five computers, not three. General Purpose Computers, GPCs. Was it "IBM AP-101/S"? Four of them run the Primary Avionics Software System, PASS. One of the five computers is selected to run the Backup Flight System, BFS, which was indeed contracted to a different manufacturer and designed independently of PASS, on purpose. The OV avionics are described as FO/FS, Fail Operational / Fail Safe. If there is disagreement or fluctuation, you can do e.g. majority vote. If one is deemed bad, the same is still possible with the remaining three (hence the "Fail Operational"). Then if a second one fails, you can't use the nominal approach anymore. -- Nuno Silva -- https://mail.python.org/mailman3//lists/python-list.python.org
