Davin News Server

In alt.global-warming Dhu on Gate <campbell@neotext.ca> wrote:
> On Fri, 4 Oct 2024 22:12:30 -0000 (UTC), R Kym Horsell wrote:
> 
>> In alt.global-warming Alan <nuh-uh@nope.com> wrote:
>>> On 2024-10-01 01:40, R Kym Horsell wrote:
>>>> In alt.global-warming Dhu on Gate <campbell@neotext.ca> wrote:
>>>>> On Sun, 29 Sep 2024 23:19:24 -0500, AlleyCat wrote:
>>>>>> Nebraska, USA: Within minutes, a single hailstorm reduced 14,000 solar
>>>>>> panels, worth millions of dollars, into a pile of toxic debris-leaching
>>>>>> materials like cadmium and lead into the soil.
>>>>> Lotta folks don't get how fragile hi-tech shit is.
>>>>> Dhu
>>>> 
>>>> 
>>>> Long gone are the days of servicing your own vehicle.
>>>> Apart from oiling some bearings you can't and sometimes are
>>>> prohibited by law from fixing your own car.
>>>> I was helping a neighbor recently and it turned out you
>>>> needed to break an old mil grade encryption to even ask it here it hurt.
>>>> 
>>> 
>>> What kind of car was that?
>> 
>> I pre-answered that. :)
>> 
>>> All cars built for sale in North America have had a standard OBD-II port 
>>> that can be read by various scanners costing anywhere from $20...
>>> ...to a lot more depending on complexity.
>> 
>> There's a theorem of computer science that roughly says
> 
> That's a "side effect" of completeness vs. consistency.  
> 
> IF you make a system "complete", that is covering all possible 
...

You're talking to someone that has been doing that stuff for 1/2 a lifetime. :)
I like to think of it in terms of self-reference.
Any system that has to look at itself will strike problems.
Can you make a machine that can diagnose its own faults?
What happens when the diagnostic part has a fault?

NASA always had an engineering policy that every mission-critical system
had  to have triple redundancy. If not 3 identical systems then
at least 3 different  subsystems had to be able to do the same job.
But the point of failure generally was the part of the system that
took the vote of the 3 identical systems to find out the majority-decision
and switch out the subsystem that was wrong.

In MV you often see the central computer start up and ask various sub-systems
to run diagnostics. The sub-systems poke some circuits and see if the
results are what they expect. If so, they send an OK back to the start-up
request from the central processor.

But the checking runs into a problem. If you have to check say 1 million
bits to decide if your sub-system is working then how do you check
they are all correct? A typical method is to use a hash. Boil the
sequence of 1 million correct bits down into a 16 bit number.
Take the hash of the million and check it against the 16 bits in PROM.
But there is a logical problem. A million bits can have "many" more
possible combinations of values than 16 bits. So it means there can
be many problems that hash to exactly the "everything is OK" code.

But the hashing idea is cheap and everyone uses it "all the time".

So it's a minor miracle these days when you put your bread in a toaster
that it cooks it for a couple mins, the toast pops up, and it's edible.