Hacker News Reader

Comment by WalterBright | Hacker News Reader

> The problem with fault tolerance is that it allows

We do that with airplanes. Think about it - you're flying at 30,000 feet, 500 mph, 50 degrees below zero, no land in sight over the North Atlantic, in a tin balloon loaded to the gills with fuel and two flaming engines.

And yet you're perfectly safe.

How did that come about? Tolerance of failure.

View on HN

The machines are designed to tolerate fault, but the FAA is designed to not let you take off unless you do a checklist that proves all the engines are working, not just the one you need for a crippled landing. So the system as a whole requires that the FAA not give in to the pressure from industry to sign off on less fault tolerance. It's a difficult issue for systemantics.

Um, there isn't a backup if the backup isn't operable before you take off.

That's their point. If the FAA didn't mandate it you can be sure budget airlines would be taking off and doing routes with one engine broken :-)

That's not a fault of the airliner design.

> you can be sure

Not likely. Who you gonna get to fly it? Who you gonna get to pay to fly in it?

> Who you gonna get to fly it? Who you gonna get to pay to fly in it?

a. people in less developed countries

b. people in said countries with less money to spend than others

this has happened before (south america in the 70s, etc), which is why the aviation industry has the regulatory system it has today

Really? You know of examples of passenger planes taking off with only one engine turning? Or of any twin engine airplane doing this deliberately (other than a test flight or desperate emergency, like the volcano is gonna blow any moment).

i was lalking about the situation of planes taking off in barely flyable/safe situations that would not be allowed by modern faa regulations, which it think is larger point that was being argued, not debating about the single engine or propeller case

sorry if we ended up talking past eachother

With wing mounted engines on two-engine airliners, there is physically no way to take off on one from other than a dry lake bed. The thrust from the operating engine will introduce more yaw on the airplane than the rudder, nosewheel steering, and wheel brakes can counteract.

Even tail mounted engines (with a shorter coupling arm to the centerline) will typically have a Vmcg (roughly, speed at which lateral control on the ground is lost with one engine inoperative) that will preclude takeoff on one (physically, not by regulations) from available runways.

Airplanes are highly standardized. Dozens and hundreds of essentially the same model are built. A few of them are built specifically to test in various ways and even crash and burn, and make sure they behave reasonably in such situations.

Civilian nuclear reactors are mostly built by a handful, rarely by a dozen. This makes learning from past mistakes and taking preventative measures across the fleet hard.

I think France has partly solved it exactly by having a small number of standardized reactors, and a number of nuclear plants which can be run in a reasonably uniform way.

> Airplanes are highly standardized.

Not really. Every one coming off the line is different. They are constantly being improved. Every part on the airplane is carefully tracked, from manufacturing lot to which airplane each is installed on. Everything is designed by engineers, not custom made on the spot by a mechanic.

Yes, French nuclear powerplants were standardized and built in batches ("séries", in French). This does not magically creates conditions for a perfect design and building process. See for example https://theecologist.org/2016/sep/29/sizewell-b-and-27-other...

Planes aren't perfectly safe (my brother was killed when SR111 crashed in 1998 after failures).

Anyone preferring not being exposed to a plane crash can abstain to travel in planes. Anyone preferring not being exposed to nuclear reactors boo-boos and hot waste has no real way to do so.

The failure points aren't always the aspects engineered by anyone related to airplane manufacture. Swiss Air 111 may have come down due to a fault/failure in wiring for its add-on entertainment system.

https://www.swissinfo.ch/eng/electrical-fire-downed-swissair...

Not just tolerance of failure. Also strict incident investigations and reporting requirements, including for "near misses"; also a strong safety culture made possible by strong unions and strict seniority-based promotion rules; also...

> a strong safety culture made possible by strong unions and strict seniority-based promotion rules

Pilots can't get ahead by cutting corners, and (to a somewhat lesser extent) it's hard for maintenance people to be pressurised to sign off on unsafe work.

Strict seniority rules means no incentive for doing quality work above the minimum.

Indeed, but also no incentive for bypassing safety checks that are redundant most of the time (which is how you get the normalisation of deviance that eventually leads to catastrophe). Sometimes that's the right tradeoff.

Could nuclear power plant operators be incentivised by using "proactive risk detection" metrics or something along those lines?

I suppose there will then be incentives to exaggerate or "invent" new risks for career advancement, but is that not preferable to the alternative?

All metrics are gameable. I think I once saw a study that suggested that every metric applied to professionals ended up having a net negative effect on actual productivity - by and large people understand their job and want to do it well, and while a metric may incentivise the few that don't, it also ends up distracting the majority.

https://en.wikipedia.org/wiki/List_of_missing_aircraft#2001–...

Look at how few of them there are, despite millions and millions of flights.

It's an absolutely incredibly good safety record. You're much safer flying across the Atlantic than driving to the airport.

Hundreds of millions in the U.S. alone, over that 18-year span.

Reply on news.ycombinator.com