Comment by jmyeet | Hacker News Reader

Comment by jmyeet | Hacker News Reader

There's a surprising amount of corner cutting that goes on with large engineering projects. Examples:

- A water slide had basic design flaws. That water slide would later go on to decapitate the son of a Congressman [1];

- The Millenium Tower in San Francisco was allowed to be built without the piles going down to bedrock [2];

- A borrowed system from an earlier rocket was never tested on the Ariane 5. An integer overflow basically caused the Ariane 5 to blow up on launch [3];

- Famously, the Challenger disaster came down to some O rings becoming brittle in the cold launch conditions and the launch was approved regardless [4];

- To counter the competition of the Airbus A320neo, Boeing decided against a full redesign of the 737 and instead just updated the aging 737 design by putting more powerful engines on it and moving them forward. This could lead to stalls if the plane angled up too much. To counter this, Boeing added the (now famous) MCAS to keep the 737 type rating [5] and didn't really tell airlines and pilots about it. But the worst part was the system had no redundancy. It relies on a single sensor. Normally for safety-critical sensors there is triple-redundancy. So for something like this you'd have 3 sensors (minimum) and there'd need to be agreement between at least two. That was an optional upgrade.

[1]: https://www.nytimes.com/2018/03/26/us/waterslide-boy-decapit...

[2]: https://practical.engineering/blog/2021/11/10/what-really-ha...

[3]: https://www.bbc.com/future/article/20150505-the-numbers-that...

[4]: https://www.simscale.com/blog/space-shuttle-challenger-disas...

[5]: https://medium.com/the-systems-engineering-scholar/inadequat...

The concept of "cutting corners" is difficult to define.

Good engineering is not about pouring as much concrete as possible into the largest hole ever made by mankind, for every bridge pillar.

It's the opposite! Good engineering is finding the minimum amount of concrete that will hold the bridge under all possible foreseeable circumstances.

Yes, sometimes the minimum proves too short, and in those times the failures are often catastrophic. But we're still looking for that minimum. We're never looking for "maximum safety with no regard for anything else". That would be impossible.

And when we push that too far, or allow economics to overrule the fact we're putting squishy humans at the center of forces thhe vast majority of them have difficulty comprehending, people die.

I'd prefer people stop couching engineering as "the quest for the cheapest we can make it", and more "lets not make an industrial scale coffin".

None of which is possible without someone who can rein in risk comfortable execs with a hard, unoverridable no.

Unfortunately, that's wishful thinking. It just doesn't work that way under any kind of profit motive.

Take a look at the third diagram in this paper titled "Risk Management in a Dynamic Society": http://sunnyday.mit.edu/16.863/rasmussen-safetyscience.pdf

Essentially, the "gradient descent" of business interests will invariably push safety design right to the edge that is deemed acceptable. Not to some "safe region" with decent wiggle room! The very edge.

This is why the "minimum amount of concrete" will eventually be the minimum amount of concrete, and catastrophic failures will occasionally occur, even in foreseeable circumstances.

> safety design right to the edge that is deemed acceptable. Not to some "safe region" with decent wiggle room!

And that is why having a safety margin is what is “deemed acceptable”. We do it with aviation and it has an excelent track record.

We did it (past tense) in aviation, up to the point where business pressures pushed things to the edge. That's how MCAS happened and 346 people died -- because due to a profit motive, the minimum number of sensors (1) were installed instead of the desired safe number (3).

This is truly the wisest comment on this otherwise very questionable thread.

It is hugely challenging - and rewarding - to try and solve a constraint satisfaction problem that balances the risk, cost, upside that are a function of the huge set of decisions made in any complex undertaking.

It is massively easier to identify flaws and demand mitigations, which you can do ad infinitum.

I mean, yes, there are a lot of designs that fail, but the vast majority of infrastructure operates reliably for decades. Most waterslides don't decapitate people, and skyscraper collapses are exceedingly rare. Launch vehicles fail more frequently but they're phenomenally complicated machines that are produced in low volumes and flown infrequently.

Reply on news.ycombinator.com