Re:10% of Firefox crashes are caused by bitflips
by gmueckl on Friday March 06, 2026 @02:30AM
(#47272064)
For safety critical systems, one strategy is to store at least two copies of important data and compare them regularly. If they don't match, you either try to recover somehow or go into a safe state, depending on the context.
[ Parent
| on HN ]
Re:10% of Firefox crashes are caused by bitflips
by d1sxeyes on Friday March 06, 2026 @02:44AM
(#47272158)
At least three copies, so you can recover based on consensus.
Re:10% of Firefox crashes are caused by bitflips
by Dylan16807 on Friday March 06, 2026 @02:52AM
(#47272211)
If your pieces of important data are very tiny, that's probably your best option.
If they're hundreds of bytes or more, then two copies plus two hashes will do a better job.
Re:10% of Firefox crashes are caused by bitflips
by d1sxeyes on Friday March 06, 2026 @07:24AM
(#47274118)
Ah, true! You just restore the one that matches its hash. Elegant.
Re:10% of Firefox crashes are caused by bitflips
by rixed on Friday March 06, 2026 @11:39AM
(#47277286)
A single hash should be enough.
Re:10% of Firefox crashes are caused by bitflips
by Dylan16807 on Friday March 06, 2026 @02:55PM
(#47280237)
Yes, but what's easier depends on layout. "Consensus" makes me think of multiple entire nodes, and in that situation you can have a nice symmetry by making each node store one copy and one small hash.
If you're doing something that's more centralized then one hash might be simpler, but if you're centralized then you should probably use your own error correction codes instead of having multiple copies.
Re:10% of Firefox crashes are caused by bitflips
by pizza on Friday March 06, 2026 @06:20AM
(#47273590)
“never go to sea with two chronometers, take one or three”
Re:10% of Firefox crashes are caused by bitflips
by DennisP on Friday March 06, 2026 @11:56AM
(#47277593)
Seems like chronometers would be a case where two are better than one, because the mistakes are analog. If they don't exactly agree, just take the average. You'll have more error than if you were lucky enough to take the better chronometer, but less than if you had taken only the worse one. Minimizing the worst case is probably the best way to stay off the rocks.
Re:10% of Firefox crashes are caused by bitflips
by robocat on Sunday March 08, 2026 @06:05PM
(#47302063)
And for breaking failures, two is way better than one! Having zero working chronometers would be bad.
Re:10% of Firefox crashes are caused by bitflips
by DennisP on Sunday March 08, 2026 @09:34PM
(#47303798)
And come to think of it, if the two chronometers are wrong in different directions, then the average could be more accurate than either of them.
Re:10% of Firefox crashes are caused by bitflips
by qznc on Friday March 06, 2026 @11:48AM
(#47277448)
In many cases the system is perfectly safe when it shuts off. Two is enough for that.
Re:10% of Firefox crashes are caused by bitflips
by Helmut10001 on Friday March 06, 2026 @04:35AM
(#47272873)
I use ZFS even on consumer devices, these days. Parity checks all the way!