Trains in Switzerland are not allowed to have 256 axles | Hacker News Reader

Trains in Switzerland are not allowed to have 256 axles

twitter.com

193 pointsedward90 comments

To save you the click through:

> This may be a great obscure fact, but it is not an example of European regulations gone mad. To keep track of where all the trains are on the Swiss rail network, there are detectors positioned around the rails. They are simple detectors that are activated when a wheel goes over a rail, and they count how many wheels there are to provide some basic information about the train which has just passed. Unfortunately, they keep track of the number of wheels using an 8-digit binary number, and when that number reaches 11111111 it rolls over to 00000000. Any trains which bring the count back to exactly zero move around undetected, as phantom trains.

From: "Humble Pi: A comedy of maths errors" by Matt Parker (https://mathsgear.co.uk/collections/books/products/humble-pi...), page 191.

edit: Added quote attribution.

So if you want to move around Switzerland undetected, commission a train with 256 axles and bam! You're invisible :P

I want to see the Snowpiercer / Speed reboot based on this concept.

Haha one can imagine this will be rapidly bred out of the species since it's such a significant maladaptation. Like dressing in Vantablack and running on a rural expressway.

To catch intruders dressed in Vantablack, you'll need one of those devices they use to find exoplanets.

Or, you know, infrared cameras :)

Or just more light, and focus on shooting the dark spots.

Or just _more_ light, and you don’t need to shoot because everything vantablack will be aflame...

What if we tried more power?

https://what-if.xkcd.com/13/

"There have been plenty of times where a hardware issue has been covered by a software fix, but only in Switzerland have I seen a bug fixed with a bureaucracy patch."

That's an amusing perspective, but it actually looks like it might be a hardware issue that cannot be patched purely with software, as it arises in the trackside devices that detect passing trains.

I hope the method for verifying compliance with this rule does not suffer from the same problem (it may be moot: a train of 64 20 meter wagons would approach 1.5 km in length; not unusual in some places, but perhaps not Europe.)

It is a bit of a stretch, but Boeing's MCAS problem might be also classed as a software issue 'fixed' with a bureaucracy patch, as the failure to make use of the available hardware to make the MCAS software fault-tolerant was 'patched' with an FAR directing pilots to the existing trim runaway procedure.

It's also a silly quote because people workaround broken software all the time.

And if you see bureaucracy as software, it's full of bugs and processes that work around those bugs.

Bureaucracy fixing software issues is quite common. One story i have heard is that in the early days of railway reservation computerization in India, there were storage issues on the servers. The solution : cut the advance booking time from 90 days to 45, where it stayed for many years.

In similar news, popular athletic app Strava has announced that they will be dropping Bluetooth support because of an issue with their code. https://support.strava.com/hc/en-us/articles/360032979131-Re...

Strava aren't fixing the bugs, just dropping the feature altogether. They even admit that other athletic apps implement Bluetooth without bugs, so it's not a platform limitation.

So sometimes people decide that an easy fix is a better choice than a "correct" fix, which would be either adding bits to the train axle counter, or implementing a way that didn't care how many axles there were- such as having a head and tail indicator for a train, and not counting the axles in between.

They obviously have a head and tail indicator already, so they don't merge all the trains together. The issue is that they overload the axle mod 256 counter with the train indicator, instead of using a separate byte.

This section is not present in the 2015 or 2016 edition of the rulebook: https://www.bav.admin.ch/bav/de/home/rechtliches/rechtsgrund... Seems like overflow detection (or more bits) were added to the counters at some point.

Yep. They routinely run fairly long trains over the Lotschberg and Gotthard lines through the base tunnels that I’m pretty certain are approaching 256 axles, but I have yet to see any with more.

Most likely the main transalpine lines now allows >256 axles while other secondary lines have not been upgraded.

Trains longer than 256 aren't the issue as the numbers roll over. The problem is trains with exactly 256 axles (or a multiple of 256), they're seen as zero.

Not an expert, but I imagine it could still cause some routing problems if a 258-axle train were mis-registered as a 2-axle train; the latter takes up dramatically more space on the network.

Considering that a train can require 2 km to break, I'm sure they keep enough space between trains that there's plenty of space for both 258 or 2 axles trains. Knowing a train is there seems much more important than its lengths (up until they decide that 2 axles trains can be much closer to another trains).

You can carry a maximum of 255 rupees in the original Legend of Zelda, and back in the day magazines commented on the "oddly specific" number.

That reminds me back when WoW was still 32 bits and people discovered the gold cap was 2 147 483 647 copper coins.

So your gold value was signed? Does WoW allow for debt by tracking negative values, or was it just an oversight?

Some people just really, really dislike unsigned ints.

I've run into programmers that think something terrible is going to happen if you use an unsigned int for anything.

Is there a popular language these days that throws an exception whenever an unsigned int overflows?

Rust in debug mode or with a compiler option will panic if a number overflows.

Huh. That doesn't sound like something people use in production code. Maybe a hardware feature is needed? I mean, it's normal with floating point to handle overflow in a more sensible way, isn't it...

There are safe operation extensions like "save add", "save mul" for numbers in Rust which would detect overflows, but you'd have a drawback as it won't return just the number (of type u64, for example), it would return result<u64> and you'd be forced to handle it.

There's also functions which people do use to specify specific behaviour no matter the release mode, where handling overflows properly is particularly important - e.g. cryptography-related code.

I think you can specify 'checked' in C# which will result in an exception on overflow.

Signed numbers are preferred. Imaging you try to buy something checking if the remaining balance would be greater than zero is pretty much guaranteed, resulting in 4+ billion instead.

Signed number (just 64bit) is the correct number type for such operations.

Pushing it off to the negatives does not solve the issue of overflow.

> Pushing it off to the negatives does not solve the issue of overflow.

it allows for comparison with errors. Working with unsigned numbers is error prone... and I started with 6502 that didn't have signed numbers.

The idea is preventing the overflow by being able to check for negatives easily. Having negative gold is likely better than leaving someone with 4billons of. For instance stuff like "max(0, balance)" just doesn't work with unsigned ones.

If the domain is known to be positive, you can use the sign bit as an overflow flag.

Something even cooler, Lineage 2 (inception times similar to wow, say 2004) had a Karma system for killing other players. Pretty much it was possible to kill anyone, without their consent outside towns. Karma required cleaning -- killing normal mobs (instead players), having karma and having unrepented sins (killed 5+ players who didn't fight back), dying - including being killed by other players resulted in gear drop. Overall it was possible to hunt such players.

There was one player who managed to kill so many new players that his 32bit karma turned negative via integer overflow and he was free from all the filthy red status.

TIL: Per the article, this 256 limitation also applies to number of users in WhatsApp conversations and the maximum block height of a Minecraft world.

Also reminds me of the maximum amount of coins you can have in runescape: 2.147B. They even created silver coins worth 1000 normal coins as a workaround for people who get to a “max cash stack”.

(Max value in java, even though the game is non longer written in Java)

The language is irrelevant; this is the maximum value of a 32-bit signed integer. Specifically, 2,147,483,647.

Java (and other languages) have other ways of representing numbers which allow larger values, but once you've encoded values into wire protocols and persistent data structures, they tend to be hard to change.

I think OSRS is still all Java, no?

I understand why an 8 bit number would be used for counting...

But that limitation makes no sense for WhatsApp user limit count. Presumably every user has an id number identifying them (maybe a long integer). You need to maintain a list of these users so your array is already way larger than 8 bits.

Can someone intuit why the 256 limit in WhatsApp?

I'm waiting for the freak accident when a 200-axle train enters a given segment, and then later a 56-axle train accidentally ignores a red light, and plop the segment is now marked as all-clear again.

Or a 200 axle one breaks somewhere and 56 axles roll back over the sensor.

You cannot "accidentally" ignore a red light. Driving over one will trigger an emergency stop unless you specifically disable this mechanism.

allowing for not just two but three trains in a given section... wow!

The thing is Switzerland is also the only country where I could imagine someone might know such a rule. Say you are working on connecting some trains, and you happen to be one car away from 256 axles. Well, if the org you're working for is SBB/CFF/FFS I could totally imagine the foreman saying "oh wait, there's a rule about this". Any other country, no way.

From personal experience people there seem to follow the rules very meticulously.

I would wager that they might be as meticulous in Japan too.

I’m glad I read an article on HN about pointing and calling[0] in the Japanese rail system as a way of preventing accidents. I take a blood pressure medication and had once or twice taken it a second time in the day, which wiped me out and made it VERY difficult to work (for the record, I don’t take it for blood pressure and it’s a very low dose, so I wasn’t at risk of my heart stopping, just very sleepy). Now I do a specific gesture, point and call out that I’m taking my pill. I haven’t accidentally taken a second one since because it makes taking my pills so much more memorable.

https://en.m.wikipedia.org/wiki/Pointing_and_calling

Ah yes, there was a popular discussion about it on HN:

Why Japan’s Rail Workers Point at Things - 959 points: https://news.ycombinator.com/item?id=14011793

Thats the one. Thanks!

For your specific issue, might I suggest getting one of those pill boxes with one slot per day? It makes it fool proof to get daily pills right, by offloading your memory to an external device.

That's a good trick. I've heard about it before but never thought to apply it to my own life. There's a lot of everyday mindless tasks that are easy to forget you've done them.

I should start doing it with shampoo when I shower.

They wouldn't need to to remember the rule, they would follow a checklist. And someone else would check that they are following the checklist.

The rule would manifest itself in some sort of control in the train design/assembly systems, which would forbid a combination of wagons totalling 256 axles. That way, the issue would never reach the shunters who actually assemble the train, as the list they receive would always avoid that situation.

If the law prohibits 256 axles, it should really prohibit any multiple of 256. A 512-axle train would similarly be erroneously detected as having zero axles.

That would be a freight train consisting of 128 units. Not likely in Europe, where the average freight train has 25 to 30 wagons, and the longest freight train seems to measure 835m, which will put you in the neighbourhood of 40 units. I'm not even sure that one would fit on the Swiss network; it runs or ran between Hamburg and Denmark, a route which features not quite as many elevation differences and tunnels.

Considering this, the chances of even hitting the lower 256 axle mark are probably negligibly remote as well.

https://en.wikipedia.org/wiki/Longest_trains gives 180 cars for the US, and longer for Australia, Brazil, and others. Cars can also have more than four axles; at the extreme you have https://en.wikipedia.org/wiki/Schnabel_car

But yes, this is probably not an issue in Europe.

This seems like a perfectly reasonably limitation in the context of european railroading (shorter trains, more frequency), when you consider that many European freight cars are 2 axle

Yes, generally freight trains are not allowed to be longer than 740m, because there would be few sidings or loops long enough to put them in. So in practice the axle limit is unlikely to be the actual limiting factor.

Previously: https://news.ycombinator.com/item?id=12071240

Shouldn't they also not have anywhere around 256, because if a 258 axle train has a couple of missed detections it'll still look like 0?

From the previous thread, some countries specific a little below 256 for a margin of safety:

https://news.ycombinator.com/item?id=12071686

I'm not sure why they're even counting past 2 (the minimum number of axles to indicate a railcar is present, at least until someone delivers self-balancing single-axle railcars) for train detection. Unless it's to estimate track wear, or for revenue collection?

Trains are typically made up of multiple carriages. The axle counters ensure that all the carriages which enter a section of track leave that section of track, and that none have detached and been left behind for the next train to run into.

To elaborate on this, I think what is going on is a section has a "counter-upper" on one side, and a "counter-downer" on the other side.

So while the train is within that segment, the value of the counter is non-zero. Once they exit, all the axles should get counted again in the opposite direction and reset the count to zero.

The problem with 256 is that the count would be at zero while you were in the track segment.

One of the primary uses of these kinds of devices is in lieu of track circuits: for a given rail block, they count the number of axles currently present by subtracting the number of axles that exited from the number that exited. Another use is to detect trains that have a failed coupling, disconnecting several cars which no longer have a powered locomotive attached to them to move them.

Looks like this leaves the issue of 256 multiples unresolved.

In theory we could still have a 512 axle ghost train.

But did you know an Oyster card in London can't have more than 16 fare zones? This is the reason why some stations on Overground don't accept Oyster cards.

There was something posted on HN years ago about a mainframe that ran the French railway pension system. There was something funny about different pay rates for rates going uphill or downhill, whether the engineer was a military veteran, etc. Does anyone have a link? I can’t find it now. Thanks.

The "fixed with bureaucracy" thing is entertaining, but it's probably a delusion that the fix is effective.

For one thing, this rule is not intuitive and people aren't likely to remember it. People have an easier time remembering and following rules whose reasoning clicks with them. It's not a recipe for success to say, "Oh, there's this one weird thing that you can't do, and it isn't going to make sense to you why, but remember not to do it."

I also wonder whether, even if someone in the industry does know the rule and really wants to follow it, there's an easy method to know whether their train has 256 axles. Is there a procedure in which axles are counted or an automated system that does it? If it's a manual procedure, is it time-consuming and is there pressure to skip that step when you're in a rush?

And, if a fallible human being should make a mistake and run a train with 256 axles, is there any protection against the phantom, undetected train scenario, or is it a silent failure?

What if Swiss trains consist of 30-40 cars at most? Then the rule is purely technical and doesn't even have to be remembered.

Very fair point. As long as that condition holds, then I guess it's a non-issue.

There are apparently trains bigger than that in some parts of the world (such as the US and Australia) but maybe conditions in Switzerland prevent that from happening? This is way beyond my level of knowledge though.

Actually, the longest regular Swiss trains seem to top out at 44 cars. This is because of passing siding lengths being much smaller than other countries such as the US. There are longer trains that are scheduled on routes with longer sidings however such as a the main north/south routes.

Most European trains rarely exceed 22 cars for the same reasons as they have to be able to be passed by express trains in stations.

Yeah, a bit of light browsing yields a 25 to 30 unit average for European freight trains¹.

https://www.allianz-pro-schiene.de/themen/aktuell/740-meter-...

Ah, it’s up from last I checked. Thanks!

but if most of the cars have 8 axles per wagon, they'll hit 256 pretty easily. I don't know how many freight wagons have that many axles, but a lot of passenger carriges do.

I am somewhat interested in railways, but I have never seen an 8 axle passenger wagon. Even 6 axle passenger wagons are not in use in Europe today. Can you give a reference of any 8 axle wagon? The longest passenger trains in Central Europe are around 400 meters. To get 256 axles there would be an axle every 1.6 meters on the whole train.

Freight trains of 1000 meters are exceptional in Europe. You would need an axle every 4 meters to come up to 256.

The axle limit is completely theoretical. The are plenty of other limits a train would exceed before.

Is 256 the upper limit, or is it just a magic number to be avoided? Is a train with 258 axles OK?

If you click through, 256 overflows to zero, which means the system thinks there's no train. 258 is fine.

> "Any trains which bring the count back to exactly zero move around undetected, as phantom trains."

The law says "equal to 256", so the higher numbers should be fine.

Shouldn't it be "multiple of 256"? Maybe nobody does >=512 axle trains.

A train with 258 would appear to be a train with 258 mod 256 = 2 axles. The purpose of the counters isn't really to count how big the trains are, just to keep track of them. So it at least knows there is a train on the tracks.

A train with 258 axles counts as a handcar:

https://en.m.wikipedia.org/wiki/Handcar

Even worse, 260 axles and you are still considered "light" engine (light engine is a term of art which means locomotive moving on its own power without hauling anything else).

Which also presumes axle-counters are perfect...

I'm surprised they didn't also ban 255 and 257!

They should do what the Norwegian Navy does, and put barcodes on the cars.

Why does Norway do it? When their ships pull into home port, they Scandinavian. ;)

That's nothing. Signed overflow in C/C++ is _undefined behavior_. Which means there's no guarantee that you'll get anything recognizable out of it. :-)

But unsigned overflow is defined. Are we expecting a negative number of axles?

The train counter isn't even software based. I just wanted to point out the example of a much more common issue that most programmers aren't aware of. Very few languages catch unsigned integer overflow as well. The only popular one I can name is Swift.

Wah -- wah -- waaahhh.

Reply on news.ycombinator.com