[kuhn]

Interesting write up. I was just watching a video yesterday which had a pretty good example of the testing aircraft manufacturers do for this scenario. It’s called a velocity minimum unstick test. Some impressive flying.

https://youtu.be/BWwUTJM3jbA

[zodPod]

"They put the plane into a stall to make sure it can recover" CAN YOU IMAGINE? lol Could you imagine being the guy who wakes up in the morning and is like "Today I get to put a plane into stall to make sure it can recover" shivers

[sjburt]

This is a normal part of pilot training. I think it's less common in large aircraft but for small planes it's not uncommon.

[mulmen]

Normal pilot training does not involve testing aircraft limits. Stall training happens in a plane with a known stall speed and behavior. The quote is about finding that limit for the first time.

[nsxwolf]

It really doesn’t feel like anything special is happening either, and recovery is quick and easy. It’s a non-event at altitude.

Spin training though, no thanks!

[ubernostrum]

Well, the nice thing about most planes is that they don't want to stall, and most of the time the correct action to recover is "stop doing things and let the plane recover". It's when you keep messing with it that you make things worse and get toward unrecoverability.

[zodPod]

I'm going to reply here because multiple people said it:

I'm aware that stall training is a common occurrence but the difference here is that you're testing with a plane that is massive and may have no nearly been tested as much before. What struck me was the "Make sure it can recover" it's not a test for the pilot's training to see if he can recover it. It's testing the plane's capability to recover. Sure the pilot could still end up recovering it if it doesn't but the sheer size of this plane is what mostly made me surprised!

[Doxin]

I'm fairly sure that the engineers have figured out a pretty good estimate of the stall speed before they send a pilot to go test it

[nunez]

Recovering from a stall is something you learn very very early into flying

[chii]

I'd say impressive commitment to QA'ing your product! I wish software in general, is tested to this degree.

[hueving]

Testing costs money. If software in general was tested to this degree, we would still be stuck in the CLI-only world of 30 years ago.

[yebyen]

Or, we would have long ago found better ways to adopt design and development strategies that more generally lend themselves to rigorous and easily repeatable testing.

[krupan]

We would have found Fred Brooks' missing silver bullet?

[yebyen]

:D

I don't know what that means, but I'm going to upvote you and go find out what interesting reading material I have missed!

Edit: As a rails dev, I know that there is no silver bullet, but I also know that some strategies are objectively better than others, and some strategies produce software that is easier to verify than others.

You can write heaping mounds of Javascript that you pile on top of each other, and pile more and more tests on top of each other and run them within a CI framework between each change (or don't), to build confidence that you don't break things that you previously guaranteed from one commit/release to the next...

Or you can write tiny functions with an obvious and wholly singular purpose, and build them into micro-services that deliver one small fraction of the whole service, and test at each layer, and build composable units of the final service that you intended to guarantee.

One strategy is easily verifiable if you keep it up without ceasing (and the other is not, but the two strategies are usually easy to tell apart.) Unfortunately it's probably true, for example, that neither strategy can ever be said to (obviate essential complexity! Fred Brooks) – to totally guarantee that the tail does not hit the ground on either takeoff or landing, which is a complex problem that probably can't actually be solved in units.

[hinkley]

We would be reusing the hell out of everything and we wouldn't have every single company on the internet building their own CMS.