How many hardware engineers does it take to change a light bulb?
None - we'll fix it in software.

How many software engineers does it take to change a light bulb?
None - we'll document a workaround.

How many tech writers does it take to change a light bulb?
None - the user can figure it out.

So in this case we have:

How many hardware engineers does it take to not crash an airplane with a faulty sensor?
None - we'll detect and avoid it in software.

How many software engineers does it take to not crash an airplane with a faulty sensor?
None - the pilots can be trained to disable the auto-trim mechanism.

How many trainers does it take to not crash an airplane with a faulty sensor?
None - the pilots can figure it out.

How many pilots who have no idea why the plane is reacting as it is does it take to crash an airplane with a faulty sensor?
None - the plane will do it for them.

(No, this isn't meant to be funny.)

"Garfield" - could be a person or a place, let's keep going
"Garfield phones" - OK, Garfield is phoning someone.
"Garfield phones beach" - maybe a public phone on the beach, or we'll get a useful continuation like "lifeguard" to which "beach" is an adjective.
"Garfield phones beach mystery" - OK, I've lost the plot here. Abstract concepts don't usually accept phone calls. Maybe "lifeguard" will still come next to rescue some sense out of this.
"Garfield phones beach mystery finally". Nope, no lifeguard, no rescue.
Perhaps "Garfield phones beach, mystery finally solved..." Totally makes sense, but I have to assume a nonexistent comma.
Maybe Garfield is a place with a beach, and the beach has a phones mystery. But then it should be "Garfield Beach phones mystery..."
Maybe there is a Phones Beach? But then what is Garfield doing? Surely nobody named a place "Garfield Phones"?

fivethirtyeight.com did a good job on the poll analysis. Somewhat simplified, before the 2016 election they were saying: Clinton is clearly ahead in the polls, but the amount she is ahead by is about the same as the typical polling error (i.e. when after the election we compare average-of-polls to actual-votes-cast.) This means there are three roughly equally probable outcomes: 1/3 probability, the polls are close to being correct, and Clinton wins. 1/3 probability the polls are significantly out, underestimating Clinton's support, so Clinton wins big. 1/3 probability the polls are significantly out, underestimating Trump's support, so Trump wins narrowly. (Of course their analysis was more subtle - they didn't just identify three outcomes and arbitrarily say they were equally likely, it just so happened that these possibilities came out to be roughly 1/3 each.)

I haven't looked at the analyses that were giving something like 98% win to Clinton, but likely they made a bad assumption that poll errors were uncorrelated - that if Trump outperformed polls in one state, that made it no more or less likely that he'd outperform in other states.

Thanks for a well thought out response, as opposed to those who take me for an evil anti-net-neutrality shill. (As stated in another response, I am pro net neutrality.)

To see if I understand correctly, I shall try to summarize. The colocation benefits are small and possibly non-existent. Designing regulations to encourage colocation is likely to have adverse effects (on competition) that outweigh the benefits. So enforce net neutrality, and let colocation fall where it may.

Is this about right?

I like the trade you describe for Akemai. It is mutual benefit without either side trying to squeeze maximum money from the other. One problem I see is that in the simple case you describe, Akemai has no incentive to use power efficient hardware. There are a variety of ways minor tweaks to the agreement could deal with this, which for simplicity or because you are not privy to them, you have omitted.

OK, substitute 'efficient' for 'smart'. I agree that is a better word.

If regulations favours inefficient solutions over efficient ones, this is not great. It might well be a price we're willing to pay (e.g. companies spending lots of money on advertising to get market share in a fixed sized market is inefficient), but it at least suggests there might be a problem with the regulations which could be fixed.

My non-net-neutrality solution was not put forward as 'smart' (efficient), just as a way in which non-net-neutrality might cause the efficient solution to come about. (I.e. I care that colocation happens, not that the commercial arrangements that make it happen are organized in any particular way.) I concede that it is not the only way in which things might go without net neutrality.

The point remains: under the right circumstances, colocation is the efficient solution. Implementing the solution produces a surplus, which can be shared in some way between the participants. This is all true with or without net neutrality. What incentives are there to implement the efficient solution, and how does the surplus get divided, with or without net neutrality?

I'm in favour of net neutrality. Your vitriol is misdirected. I just want to know how incentives work in various systems.

I have a general net neutrality question.

Consider that 1000 of my local ISP's customers want to watch a hot new Netflix show. My ISP is 1000 km from the nearest Netflix data center. The dumb solution is that 1000 customers sent requests to ISP who sends them 1000 km to Netflix who sends the show 1000 times over the backbone connection. The smart solution is that Netflix colocates a server in my ISP's small local data center which they send the popular shows to just once over the backbone, and this server sends it to the 1000 local customers.

For the smart solution to happen, there have to be incentives for Netflix and the ISP to do it. Without net neutrality, it could work: ISP gets to advertise that Netflix is 0 rated (or 0.5 rated or whatever) towards customer data caps, and benefits from being more attractive to customers and not paying for so much backbone data. Netflix benefits by not needing so much internet backbone. Customers benefit obviously, at least in the short term. (Customers may suffer in the long term through lack of competition.) Would-be Netflix competitors are very unhappy. Possibly money changes hands between ISP and Netflix to make this work, although I'm not sure in which direction.

With net neutrality, the ISP can't offer reduced rating on Netflix data. How do the incentives work in this case? The great reduction of data going over the backbone should provide savings, but who was paying this cost in the first place? Does the ISP want to pay Netflix to colocate a server, or to charge them for it?

I've long wondered why Boeing doesn't make a clean sheet replacement for the 737. For a plane sold in such vast numbers, the case for a clean redesign is much easier to make. My understanding is that currently they're looking to a 757 replacement/A321 competitor (https://en.wikipedia.org/wiki/Boeing_New_Midsize_Airplane) as their next clean design, with a 737 replacement possibly after that.

Back in my day, we waited 1997 and then worked excessive hours in a panic for three years. If it was good enough for us, it should be good enough for you.

I actually had a 100% genuine post-Y2K bug that I needed to fix. I was one of the main people in charge of source control (which was SCCS) in our company at the time. On 2nd Jan my on-call cell phone rang (while I was watching Sleepy Hollow in a movie theatre - oops!) A few rather keen developers were working, despite it being a holiday, and were getting weird stuff happening in SCCS. It turned out that the server they were using had SCCS via someone just copying over the binaries, rather than correctly installing them, so when it got its Y2K-compliant OS upgrade, those binaries were not replaced, and were not Y2K-compliant. I was able to diagnose the problem, patch the dozen or so source code commits which they'd made with the bad binaries, and call in a sysop to install the correct binaries. (I didn't have root on that server.)

Ditto. I felt smugly superior to be using little-endian DD-MM-YYYY which was so much better (consistent) than the USA's middle-endian MM-DD-YYYY format. Then starting in 1988, my MSc thesis was on an experiment with many Japanese collaborators (they had the money, we had the supernova) and I saw them use YYYY-MM-DD and I was an instant convert. We're already big-endian in our decimal notation, so dates should be too. And in YYYY-MM-DD, chronological order and 'alphabetic' order are the same.

Yes, saves the airline a little bit of money to not have the passenger on board. A number I came across was that for a typical jet airliner each extra 100kg of payload increases fuel burn by about 3kg/hour. I think this number was in the context of long haul flights. Short haul and turboprop aircraft would have somewhat different numbers. I think I saw this number 5 to 10 years ago, so more recent higher efficiency engines likely have a smaller number.

This clearly has consequences for airlines deciding how much fuel to load. They have to take extra fuel beyond what they expect to burn to get to the destination, to allow for contingencies (and because they legally have to.) It is something like an insurance policy: loading an extra 10 tonnes of fuel for a 10 hour flight will cost you 3 tonnes of extra fuel burned, whether you use that reserve or not. Usually you won't, but sometimes it will save your bacon and avoid having to divert to an unexpected airport, and then having to perhaps get hotel rooms for all the passengers and rebook them to their expected destination, plus annoying passengers and having them perhaps never fly with you again. Usually airlines will load the legal minimum (from memory, something like fuel to fly to destination plus fuel to fly from there to the further of the two alternate landing airports plus 30 minutes) but if in-air delays seem likely (e.g. bad weather) they'll load more.

I had the impression there was some sort of feedback mechanism in how much bitcoin you get for mining a block. TFA doesn't talk about this. This would mean that if your mining costs are substantially higher than others, you'll be unprofitable no matter what the bitcoin price is doing.

Perhaps it is just a supply/demand thing? The more miners there are, the lower the chance you get to mine the block, so the lower the expected return you get in bitcoins. Then for a fixed bitcoin reward, the number of miners will bring the expected return in equilibrium with bitcoin value.

Could someone with knowledge please talk about feedback mechanisms?

I think nature would rather strip oxygen out of the water rather than the chlorine, so you'd end up with oxygen and dilute hydrochloric acid.

I think nature would rather strip oxygen out of the water rather than the chlorine, so you'd end up with oxygen and hydrochloric acid.

I think nothing. They need a liquid electrolyte around the Na electrode, and it can't be aqueous because that would react violently with the Na. So they have an organic electrolyte into which the Na can dissolve, and then it passes through a membrane into aqueous solution where the rest of the reactions happen.