[logifail]

> Physical access can generally circumvent any security precautions, but it seems like there are some fairly simple schemes that would work well to prevent a disturbed captain

I can't be the only one thinking that in the situation where we postulate losing trust in the captain, whether tracking is switched on or not is a relatively minor issue, compared with - say - the captain deliberately flying the aircraft into a mountainside.

[ozr]

You're definitely not wrong. Most civil aviation authorities I'm aware of have some profound issues with their policies around pilots and mental health issues. Fixing that is more likely to prevent the issue in the first place. That's closer to the root cause and something that shouldn't be ignored.

But parallel to that (and unfortunately-but-most-likely in lieu of): we simply shouldn't be losing large aircraft and having no idea what happened.

[bombcar]

The biggest issue with all policies is the same with red flag laws - and even attempted mention of potentially having an issue grounds you and you lose your job; this makes the issue that much worse.

[ipython]

Yes. This is dangerous because pilots are afraid to disclose depression, anxiety and other mental health conditions for fear of losing their livelihood. For example, the FAA automatically considers an ADHD diagnosis disqualifying for airline pilots: https://pilot-protection-services.aopa.org/news/2018/februar....

This results in a “don’t ask don’t tell” situation which can be even worse.

[logifail]

> we simply shouldn't be losing large aircraft and having no idea what happened

Maybe it doesn't nearly happen often enough to make this any kind of a priority?

[ozr]

Why pay for the recorder at all then?

Sub-total loss is presumably much more common and it’s useful there, but the adjustments to prevent it being disabled/make it more durable would be helpful there, too.

[sokoloff]

From the article: All of the preparation is, inarguably, effective: Never has a recovered black box been so badly damaged that it yields no usable data.

That seems to me like they might be at a sufficient level of durability already.

[ozr]

I worded that poorly; durability in the sense of being resistant to attack by bad actors, not in the event of a crash.

[logifail]

> being resistant to attack by bad actors

The recorders already to work really well for the purpose they're designed for.

The "bad actor in cockpit" plus "unexplained disappearance" scenario, for which they're not designed, simply doesn't happen often enough to be worthwhile worrying about.

Short version: flight recorders aren't broken, they don't need "fixing"

Long version: imagine you were to propose a major redesign to your employer's systems to attempt to diagnose unaccountable failures which based on historical data happen 0.00000018% of the time[0], what would be the response?

[0] back of napkin maths based on https://news.ycombinator.com/item?id=36284912

[Tabular-Iceberg]

Presumably because the incidence of pilot murder-suicide falls below the threshold and genuine accidents fall above it.

What’s the ratio like, 1:10000?

[bombcar]

Which has occurred - perhaps if we actually want to try to control for “insane pilots” there has to be some way for someone on or off the plane to force it into autopilot with cat 5 autoland at the nearest airstrip.

[ipython]

… which then extends the critical flight controls security surface area to entities on the ground that have no vested interest in my safety? No thank you, I’ll take my chances with the flight crew which in 99.99999% of cases are also interested in arriving home safely.

[sokoloff]

There are at least 7 commercial passenger flight suicides in history (plus the four on 9/11, though that's arguably a closed vector now, and some others on non-revenue flights). There are around 100K pilots working today in commercial aviation [not all Part 121 (airline)], so perhaps 500K, and surely less than 700K, total over the course of time. If we posit a successful suicidal action rate of 7 in 700K, that's only "five nines" over the course of their shortened career or 99.999%, but could become seven or even eight nines on a per-flight basis.

I'm 100% with you on the desire to not have ground-link control (and, for me, to keep two crew in Part 121 operations); I just figured I'd estimate the math.

[ipython]

I had a feeling someone would try and do this calculation. :) I would calculate it instead based on the number of flights rather than the number of airmen. It doesn’t make any sense if the units for the numerator are “number of successful suicide flights” and the demoninator is “number of airmen”.

I would not count the four flights in 2001, personally. Those are not suicide by flight personnel.

If you look at an average of say 25m flights per year since the year 2000, then you would have about 550mm flights. Even if you include the 4 tragic flights of 9/11 in the numerator, my 7 nines are very conservative.

[WalterBright]

The flight crew always arrives at the scene of the accident first.

I've read that pilots sometimes remark that their focus is on saving their own asses, and if they do that, the passengers will be fine.

[kashunstva]

> force it into autopilot with cat 5 autoland at the nearest airstrip.

I think CAT III C ILS have the lowest (theoretical) minimums at 0/0, although I don’t think that designation is actually used; so let’s just say CAT III in general has the lowest minimums.

[bombcar]

You're right, I should have said CAT 5e (where they run a ethernet cable to the plane).