Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Delays in Boeing Max Return Began with Near-Crash in Simulator

bloomberg.com

Post reply on HN
Print view View original

121–130 of 199 posts

Re: Delays in Boeing Max Return Began with Near-Crash in Simulator

#121

Post by MFLoon »

zbrozek wrote: post #76
zaroth wrote: post #12

The can of worms has been officially opened, and this may turn out to be a case study in how bad software and outsourcing your centers of excellence can bankrupt a company. If TFA is to believed, Boeing has embarked on a project to take what is currently a master-slave failover design and jury-rig it into a master-master real-time system. This is based on bad performance in the simulator—not because of MCAS—but while…

"Dual-dual" is a thing in this space. There are safety-intent microprocessors which have a couple of cores running a couple of cycles out of sync with a delay and comparison on registers. If they disagree after that delay, the processor concludes it's insane and you can take action based on that hardware-level signal. It's not my favorite solution for every application, but it's a good tool for a handful of situation…

I'm confused as to what Boeing is changing then - your description of 'dual-dual' systems sounds like what is happening with the master/slave failover system already in place. It sounds like what they're changing is to have the two systems cross-check each other in real time, rather than have the in-band delayed core be the sanity check for each system, which seems like a big difference because it means that there are now disagreement scenarios whereas an in-band sanity check always is just checking one system against one set of invariants and thus can't disagree with it's own observations. Or am I misunderstanding?

Re: Delays in Boeing Max Return Began with Near-Crash in Simulator

#122

Post by perlgeek »

kyleblarson wrote: post #93

Possibly stupid question but will they be able to apply these fixes remotely? Or will Boeing have to fly engineers / mechanics to every grounded 737 Max in the world?

I could well image a middle ground for software updates: patching needs to be done by ground crew (and while not in flight, obviously), but not by Boeing engineers, just by regular maintenance crew.

But the article reads as if new hardware was necessary as well, which might change the picture.

Re: Delays in Boeing Max Return Began with Near-Crash in Simulator

#123

Post by anticensor »

Xixi wrote: post #84
xvf22 wrote: post #41

Earlier quoted context omitted.

Four engines are generally way more inefficient than two big ones. That's why they're being phased out. A340,B747,A380, BA146, etc.

Contrary to popular belief, quad engine planes are not inefficient [1], at least not in terms of fuel consumption (we are talking 1% to 2% difference). In terms of dispatch reliability however I think they are doing somewhat worse than twins, as they are more likely to have one engine out of service. And it not a small problem: a plane flying revenue passengers is an asset, but a plane not flying is a liability. At l…

One engine out situation is called for while planning the flight.

Re: Delays in Boeing Max Return Began with Near-Crash in Simulator

#124

Post by heisenbit »

zaroth wrote: post #12

The can of worms has been officially opened, and this may turn out to be a case study in how bad software and outsourcing your centers of excellence can bankrupt a company. If TFA is to believed, Boeing has embarked on a project to take what is currently a master-slave failover design and jury-rig it into a master-master real-time system. This is based on bad performance in the simulator—not because of MCAS—but while…

If what we read is true this is then the second time that Boeing is trying to fix hardware shortcoming with software in a hair raising manner.

Re: Delays in Boeing Max Return Began with Near-Crash in Simulator

#125

Post by heisenbit »

jeanvalmarc wrote: post #100
lisper wrote: post #97

This article is incoherent: > “It’s really complicated,” John Hansman, an aeronautics and astronautics professor at the Massachusetts Institute of Technology who is not involved in the repair, said of revising aircraft software. “It totally makes sense why it’s taking longer.” But then... > While the fix became more complex and politically charged after the second accident -- the crash of an Ethiopian Airlines jet on…

They're talking about two different changes: "bunch of grad students"="patch MCAS to not auto-crash" "really complicated"="redesign entire compute architecture to be multi-master and failure tolerant"

"really complicated"="make an architecture usually done with three computers work with just two which were incidentally architectured and dimensioned for a different job. Get this done in a lot less time than usual and btw. you don't have the usual freedom to discuss constraints one usually has at this point.".

Re: Delays in Boeing Max Return Began with Near-Crash in Simulator

#126

Post by ethbro »

MFLoon wrote: post #121
zbrozek wrote: post #76

Earlier quoted context omitted.

"Dual-dual" is a thing in this space. There are safety-intent microprocessors which have a couple of cores running a couple of cycles out of sync with a delay and comparison on registers. If they disagree after that delay, the processor concludes it's insane and you can take action based on that hardware-level signal. It's not my favorite solution for every application, but it's a good tool for a handful of situation…

I'm confused as to what Boeing is changing then - your description of 'dual-dual' systems sounds like what is happening with the master/slave failover system already in place. It sounds like what they're changing is to have the two systems cross-check each other in real time, rather than have the in-band delayed core be the sanity check for each system, which seems like a big difference because it means that there ar…

You're forgetting there's a third system: the pilot.

If the majority of computer aids turn off, the plane doesn't simply drop out of the sky. It reverts back to a more complicated, but human-controllable, vehicle.

From the article (and who knows how much Bloomberg understood) the active-active sounds more like recognizing, not recovering from, errors.

Which isn't the worst idea. Pilots being, well, pilots.

Re: Delays in Boeing Max Return Began with Near-Crash in Simulator

#127

Post by nikanj »

drawkbox wrote: post #98
zaroth wrote: post #12

The can of worms has been officially opened, and this may turn out to be a case study in how bad software and outsourcing your centers of excellence can bankrupt a company. If TFA is to believed, Boeing has embarked on a project to take what is currently a master-slave failover design and jury-rig it into a master-master real-time system. This is based on bad performance in the simulator—not because of MCAS—but while…

Classic case of hitting the date but crunched out with technical debt and buggy software, a tell that management/finance/marketing is driving the decision over the engineer/designer/product people. People forget a project was late if it succeeds, they never forget a project on time that failed because it was rushed, buggy and not polished production. MBA/business and marketing need to go back to letting the product p…

> People forget a project was late if it succeeds, they never forget a project on time that failed because it was rushed, buggy and not polished production.

#notallMBAs, but the surest way for your product to fail on the market is for your company to go bankrupt before you’ve polished it enough.

In this specific case it was about greed, but sometimes MBAs push for release, because according to the engineering teams the product is never truly perfect

Re: Delays in Boeing Max Return Began with Near-Crash in Simulator

#128

Post by thfuran »

SQueeeeeL wrote: post #106
uxp100 wrote: post #104

Earlier quoted context omitted.

Is that true that CS students don't take an ethics course? I have a 10 year old SE degree and I took 3, only 1 of which was engineering specific. (I took a philosophy ethics course, which really isn't that relevant to issues like these anyway, but also a course that was specific to ethical issues in computing.) OTOH, I'm skeptical those courses change behavior, and in this particular case others in comments are sayin…

In my opinion courses wouldn't change behavior, modern computer science is designed to infantalize coders and make them interchangeable. Think of all the effort startups put into marketing themselves as "COOL TO WORK FOR"/"WE HAVE BEER ON FRIDAYS". These reinforce the idea that you are small and need management to tell you exactly what to do everyday, because you're a dumb idiot who spends all day watching anime and…

Given

>marketing themselves as "COOL TO WORK FOR"/"WE HAVE BEER ON FRIDAYS",

"the idea that you are small and need management to tell you exactly what to do everyday, because you're a dumb idiot who spends all day watching anime and playing video games" does not follow.

Re: Delays in Boeing Max Return Began with Near-Crash in Simulator

#129

Post by ethbro »

davidhyde wrote: post #5
akersten wrote: post #2

These should not fly for a decade. They're rewriting the avionics from scratch. This should require a years-long process of testing, external auditing, and approvals. I'd even go so far as to say in flight, we should require competitor review. That's right, no trade secrets. If you want your Hello World up in the air, you better let your competitor vet the source code. This is an engineering boondoggle and an embarra…

Agreed, the MAX was only designed to beat the competition to market with the least amount of re-certification and re-training. That gamble (with Boeing's reputation, the FAA reputation and most importantly, people's lives) did not pay off. They won't get away with rushing though certification of a complete rewrite of their flight software. The world is watching now. Time to cut losses and junk the MAX entirely.

I'm not sure why this keeps getting repeated.

Boeing certainly bears culpability for failing to execute, and whatever terrible lapses in certification were made to get there.

But why were they doing a dumb thing in the first place?

Because American Airlines asked-told them to. (Because they didn't want to have to pay to retrain pilots)

Boeing deserves to have substantial portions of management jailed over this, but there's blood on the legacy 737 operators' hands as well, for asking them to do it in the first place.

Re: Delays in Boeing Max Return Began with Near-Crash in Simulator

#130

Post by salawat »

danso wrote: post #20
Tomte wrote: post #13

Earlier quoted context omitted.

> a simulated edge case involving "gamma rays from space"? Soft-errors are a prime concern when designing safety-critical systems, both in hardware and in software. It's not some outlandish edge case. Incidentally, while soft-errors in an airplane are probably primarily particles from space, on the ground we're still sometimes talking about it, but in reality it's mostly chip packages that are low radioactive and emi…

I don't disagree that this is an important thing to test for. I question why this is the reported explanation for why Boeing now decided that the computer systems needed an extensive overhaul. Ostensibly, "gamma rays from space" affecting flight control computers would be an issue that affected Boeing's planes long before MCAS. But it's not the gamma rays that are the reason for the systems redesign, it's the fact th…

Also, there is a regulation that requires no design have a single point of failure capable of resulting in a catastrophic outcome.

That's why they have to rearchitect the system. If those bit-flips happened to the Flight Computer in command, they could lose the plane. Therefore, redundancy must be built in.

It wasn't originally evaluated as posing a catastrophic outcome in the safety analysis, which is why no one paid it any mind.

Post reply on HN