[sofixa]

The MCAS failure was a multilayered one. The engineers who designed it, software engineers who developed it, QA, testing, pilots, managers who approved, managers who instilled that work culture, etc etc are all responsible and should all have been fired and sentenced to prison for their negligence and dereliction of duty.

[obscuretone]

If everyone is a little bit responsible then no one is fully responsible.

It’s like dieselgate. Or the Pornhub chaos.

The VP’s tell the directors to get it done. The directors get it done.

There’s a liability force field and those that protect and serve are elevated.

[lupusreal]

In the aerospace industry, and other safety critical industries, they use the "swiss cheese" model for risk analysis. For an accident to get through all the layers (slices of cheese) meant to stop accidents, you need the holes in all of those layers to line up in just the wrong way. When that happens you need to investigate and fix each layer of cheese, not focus on the single layer of cheese on the top and ignore the rest. You will never build reliable safety critical systems with the mentality that only a single point of responsibility needs to be identified and corrected.

https://en.wikipedia.org/wiki/Swiss_cheese_model

[obscuretone]

I've written safety-critical firmware. You'd be surprised :D

[immibis]

The phrase "accountability sink" was brought up recently in the context of AI.

[lnsru]

This thing with duty feeling is hard. They all feel they did their best. Exactly as my precursor who spend 17 years in the role and did way below bare minimum. No single comment in the code. Last project does not work at all, but was approved as finished. Laziness and ignorance are everywhere. As long as managers accept it there is no way to stop that. Maybe I should stop drawing flowcharts in ascii art manner. Nobody will say “thank you” for that anyway. Nobody will give me a raise for that. It might help the next guy in my chair, but should I care about him?

[sofixa]

> This thing with duty feeling is hard. They all feel they did their best

If you work in aerospace and your "best" is non-redundant inputs from a sensor known to fail, you're not good enough for aerospace and never were. I know not to do that and I'm just a former SRE who also likes air crash investigation. My homelab has more redundancy than a critical system that could by design crash an airplane.

And people at Boeing knew this. That's why they hid the system's existence from everyone and lied to the FAA, who also utterly failed at doing its job.

[immibis]

I heard this was done on purpose because they'd need a lot more documentation if the system was important, and having redundancy would tip off the FAA to its importance.

[consteval]

The 737 MAX was a poor, cheap design from the start. It was a cost-cutting plane and compromises, like MCAS, had to be implemented to make it work.