I would say that engineer may have a perfectly valid reason to do it, and it may be even a better overall decision, as was indeed in this case, preventing fuel pipes in the cockpit as I understood from the article.

What is often lacking is testing of significant changes before giving final result to the user. And that, together with writing documentation and training is often deemed uninteresting and boring by engineer rich teams. "Community will fill up gaps" is often heard. Well maybe. And maybe community will write an accident analysis after something bad has happened.

This highlights the importance of SOPs, training, checklists etc.

What's the value of a unit-less fuel gauge? Why not a warning light or two per tank and / or a calibrated fuel gauge?

If you want to move the fuel hoses out of the cabin (good plan!) move the fuel hoses out of the cabin but leave the interface in place and make it electrical; put the mechanical switch in the difficult to find place and update the SOP to include "if the remote fuel switch fails use the mechanical switch in this difficult to reach location"

Engineering around edge cases is hard. Which is just simply saying "engineering is hard". Doing things without addressing edge cases is just craft.

This kind of thing is where experience makes the difference. You're not going to learn these kinds of things in college.

My college engineering curriculum dedicated an entire class to the topic of failure modes, edge cases, testing, SOPs, etc., and it was something professors in multiple departments took pains to teach us (and warn us about)—even in the humanities and arts. Hell, even just putting units on everything was something my teachers drilled into us since grade school. That education has informed every aspect of my professional practice.

Now that I am old, an important part of what I do is teaching others what I've learned, because if experience has taught me anything, it's that people don't just magically figure things out. Formal education and mentorship are as vital to my younger colleagues' success as they were to mine.

Sounds like a good class.

The testing bit is kind of huge. With most computer UI usability is fairly easy to test and, obviously, failures far less consequential.

I also wonder if the builder also made a faulty assumption that the plane would be flown mostly by experienced pilots in which case they may not have done a great job anticipating edge cases.

> With most computer UI usability is fairly easy to test

If that's the case, why are computer user interfaces generally so bad?

Because a ton of software shops don't care about UI testing. They don't budget time or money for it, or they don't even know they need to be doing it.

Some user interfaces are bad by design, ala dark patterns.

I probably answered that in my second bit. There’s little (perceived) downside risk in “bad” UI. So not enough people do it.

I don’t think it’s a fluke that a lot of site optimization tools are focused at e-commerce apps and sites where the stakes are a bit more tangible.

John Denver WAS an experienced pilot. Experience can only go so far to make up for the unexpected.

True, that being said though, he had only gotten the plane the day before and practiced a few touch and go landings with it when he got it.

This was his first "real" flight in this plane.

So how is anyone supposed to gain experience in a kitplane that kills the inexperienced?

Only fly planes you have flown before?

At college one professor used a particular engineering class to weed out students from the engineering program. He intentionally made it so challenging it was considered its own prerequisite.

I found the practice abhorrent even in that context.

Agreed. When my girlfriend was on a pre-med track in college, I distinctly remembered that she graduated with her biology degree knowing the diameter of a mouse hair cell but not knowing what the liver did. She graduated sumna cum laude, but did not have a good systemic understanding of the body at all. And the reason was that premed programs were designed to weed out the kind of people who couldn’t memorize vast quantities of information.

> Only fly planes you have flown before?

Come on. I can hardly make the problem more obvious, but I'll try:

If people did that, nobody would fly any plane ever. You have to fly for the first time at some point.

When there's no simulators or co-pilot seats like for the kit plane in question, you literally have no other way to gain experience.

This is the equivalent of job offers asking for 10 years experience in a technology that's barely a few months old.

Exactly!

I was speaking ironically

As my comment noted, I am not a fan of making something its own prerequisite.

My bad. It appears I'm the one who should've read more carefully.

You don't just get to override Burt Rutan's design because you have a "perfectly valid reason to do it." For something like an aircraft you need to prove beyond a reasonable doubt that your way is better and you've thought of everything he did, or build it to spec.

It's an experimental amateur-built aircraft. The builder gets to do nearly anything they want to do (usually) even if it's a bad idea, and most of them have minor changes from the original plans.

The design change was possibly an improvement when the only user was the builder who was intimately familiar with its design, but ended up being a disaster when someone else tried to use it under stress.

You do, and this is why the FAA makes you put EXPERIMENTAL in big letters in the cockpit. If you want proof beyond a reasonable doubt that the plane is safe, get a certified aircraft.

IMO the FAA strikes an excellent balance between allowing people to make crazy flying contraptions that can be dangerous, yet also keeping the civilian flying population safe. One place where there can be some unanticipated risk though is, like in this case, a homebuilt aircraft that's sold to a non-builder.

In any case though, in my opinion, the primary cause of the accident was not the fuel selector, it was John Denver's lack of preflight. He took off with very little gas in the plane, and virtually none on the tank he had selected for takeoff.

If you want to make a change, you must solve the WHOLE problem, not just the part you want to solve.

You need to make that valve easy to operate, operate in a sensible way (flip right to the right fuel tank, not left tank, have a 'both' setting, etc.), and make it ergonomically work.

In this case, the ergonomic issue could have been solved with a "dead pedal" right next to the right rudder. This is a fixed surface adjacent to the pedal in the cockpit wall you can put your foot on and press. There is one to the left of the clutch in most well-setup sportscars. With this, the pilot could press and reach around without any input on the rudders.

But, this "builder" couldn't be bothered to actually solve the problem and he killed a bunch of the pilots using his update, including John Denver. Criminal, if you ask me.

It's not a misunderstanding of the original design. It's ignoring it completely.

It's someone unqualified to be making engineering decisions absurd level of optimization for one rare (to the point of being nonexistant) risk, to the point that they made a valve that is part of near constant use in many planes...nearly impossible to see or reach.

Alternating what tank you draw from is part of normal aircraft operation because otherwise you end up with balance issues. Making that valve require contorting oneself and fully deflecting a flight control surface is so fucking asinine it defies belief.

That's not true at all. My airplane, which is a Part 23 certified, factory-built aircraft that was made by Grumman-American, has numerous fuel lines running through the cockpit, and none of them have any fire protection on them, not even firesleeving. They are made from aluminum and copper, which are not considered intrinsically fireproof. That's also the case for every Piper and Cessna aircraft that I have worked on, which are typically older CAR-3 certified designs.