[solarkraft]

The touchpad's feedback is implemented very low in the firmware, so it almost never fails to respond. The device needs to be catastrophically out of power or catastrophically locked up. Both states are not easy to achieve.

[endemic]

Anecdotally, it’s happened to me multiple times. A good indicator of when the machine needs a hard power cycle.

[f1shy]

I had it 2 times in 7 years os continuous use of the Macbook. It is very rare indeed.

[exallocatepool2]

saw your comment: https://news.ycombinator.com/item?id=34164748 and now I'm curious which manufacturer it is? is it a manufacturer of expensive german cars?

[f1shy]

I would like to keep that not public, I really fear they suing me.

I know many people working in car manufacturers. There the situation is much much worst, as they do not develop, but do the "specification" -- I see a big inclination to think that "doing" anything is pretty easy, what is really difficult is "specify" it.

One friend of mine works in such a company, and, just as example, they specified that all internal communication in the car must be IPv6 (about 10 nodes, so no need for a lot of addresses) -- Maybe they think "more modern, more better". The problem is, the whole internal communication is CAN over IP, meaning 99% of packets are 8 bytes payload, so using IPv6 adds terrible overhead. To the outside world IPv4 was specified, for "compatibility". The interfaces where specified with 2 to 10 Gbps, to "keep latencies low".

More interesting are the discussions of the need of PTP over NTP. Because for some reason I hear "NTP 1 second accuracy, PTP 1 nanosecond accuracy" constantly. They nice thing is: other than airbag, injection and ignition, all other things in a car can tolerate with ease 100ms to 300ms latencies. Even breaks con live with little less than 100ms. So there is no real need for PTP... but "more modern must be more good"... :)