Molly guard in reverse | Hacker News Reader

Molly guard in reverse

unsung.aresluna.org

222 pointssurprisetalk93 comments

https://en.wiktionary.org/wiki/molly-guard

In DevOps (and Lean, TPS) the more advanced form of this is the Poka-Yoke (https://en.wikipedia.org/wiki/Poka-yoke). Poka-yokes don't just add safety, they also guide the human away from making a mistake.

The canonical example is the automatic shift knob in a car. The shift knob is designed to 1) prevent you from accidentally shifting all the way back into reverse without pressing the shift button, and 2) prevents you from leaving park or neutral without depressing the brake pedal. This way you don't damage the drivetrain or accidentally cause the car to roll forward/backward.

Poka-yoke is a form of defensive design (https://en.wikipedia.org/wiki/Defensive_design). For a beautiful example of defensive design, see the average electric kettle. If water boils over the top it won't short the device, if it boils dry it'll stop operating, the handle and body are plastic to prevent burning yourself, the handle is ergonomic to make carrying 1.5L of sloshing boiling water not cause you to spill it, the cord is detached from the kettle so you don't yank the cord and spill the boiling water, the switches are located on the bottom away from hot steam, and the lids usually lock while in operation, again to prevent damage from spillage or steam. It's the simplest and safest possible way to boil water, and it's $20.

My favorite example of poka-yoke is when the pieces and hardware in build-it-yourself furniture kits won't fit anywhere except the correct places: two screws only have the same width if they're interchangeable, wood bars refuse to go in unless facing the right direction, etc.

The example that comes to my mind is lockout tags. [0] It usually means temporarily jamming up a specific control marked as the lockout/ignition/energizing control while you're working on some big and gnarly machine. There's a bunch of regulation around the specifics of what that control has to prevent if not activated/lockedout, but usually it's a dirt-simple breaker switch or hydraulic valve, controlling whatever the main source of energy into the machine is. The ones with holes are for padlocks that everyone will lock padlocks onto so you have a count of who's still "down there".

If you ever URGENTLY needed to start a machine, and you knew it was safe to do so, the average shop gremlin could always break the tag and start it since they're normally made of craptacular plastic or thin sheet metal... but it's easily enough friction to make you rethink what you're doing. Never known anyone that's ever had to break a tag like that.

[0] https://en.wikipedia.org/wiki/Lockout%E2%80%93tagout

There's a great example I found a while back when I replaced the fuse box in one of my Range Rovers.

It has seven plugs (there's space for eight) each of which have space for eight pins. The plugs are identical - almost. They're different colours, and they have a T- or U-shaped pin that fits into a hole in the appropriate flying socket on the engine bay wiring harness. The pins are rotated for each plug. [1]

There's no way to fit the fusebox the wrong way round in the engine bay because it has three mounting holes with odd spacings, and one has an angled slot for a bracket that holds a coolant pipe which definitely wouldn't fit if it was wrong.

There's no way to fit the sockets in wrong even without the pegs because the wiring harness only allows them to line up with the correct plugs.

Even the three high-current screw terminals that feed the body ECU under the driver's seat have got little lugs sticking out so you can't mix them up, although since they're all unswitched feeds fused at 60A it kind of doesn't matter.

There are a lot of nice little bits of design like that. Shame they didn't extend that to the ignition coil connectors on later V8s, which are the same for both pairs of coil packs. See if you can guess what causes a lot of "crank, no start" faults when people have been in at the back of the engine.

[1] https://bparts-eu.s3-eu-west-1.amazonaws.com/images/62538/bi...

Funny you say that. As I was using it at a hotel, I was just reflecting on how poorly designed the average kettle is, with the stationary handle on the top and a spout guard: https://stock.adobe.com/video/hot-cup-tea-kettle/189747353

Once the water is boiled, you flip the guard, and get your first scalding. Pour the water and get a nice splash of steam due to the fixed position of the handle as it rises up past your hand. And refilling the kettle is yet another opportunity to get scalded, as the handle gets in the way of your efforts to remove the lid.

Electric kettles are much better designed, and are what the parent comment is talking about. This [0] is the canonical white-labeled kettle that you can have for 30 of your hard-earned canukistanian kopeks, and are the average kettle around here. No scalding since the lid's handles are deep enough that you can open it on an angle to direct the steam away from you. Big handle on the side instead of the top. Switch on the bottom away from hot metal casing. And the kettle just lifts off the base, so you're not futzing around with a cable while pouring hot water.

No idea why these aren't common everywhere.

[0] https://www.canadiantire.ca/en/pdp/black-decker-cordless-ele...

Thats certainly an iconic kettle look, but I'm not sure I'd call it the average kettle. I don't think ive ever seen someone actually use a kettle like that. Maybe many decades ago.

I use one much like this four or five times a day, and have none of the problems described. I fill it through the spout though, so never have to remove the lid.

I haven't seen a stove top kettle since the 1970s. And they were an old design then. I have never seen one in a shop either. A typical electric kettle has the handle at the back and spout at the front. There isn't really anyway to come into contact with steam in normal use. They are cheap, safe and fast.

There's a great piece of software called "molly-guard", which intercepts calls to "poweroff" and "reboot" and similar. It checks if it's being invoked via an SSH session, and if so, it asks you to type the name of the system you're shutting down. That way, you never accidentally shut down a remote server when you meant to shut down your own system (or a different server).

Another fun one is disabling the network interface on a remote server. An acquaintance did that by mistake on a cloud VM running some core services, and the cloud provider had no virtual console for some reason. Ended up having to write off the VM and restore from backup. Fun day at the office.

Long ago, I succeeded once to cut my own access through SSH to a remote server, after some firewall changes. That of course has required a long trip to the server, for physical access.

However that was good, because after that I have always been extra careful at any changes that could affect the firewall in any way. (That is not restricted to changes in firewall rules, because there are systems where the versions of the firewall program and of the kernel must be correlated, so an inconsistent update may make the firewall revert to its default state of denying all connections.)

I can warmly recommend the nohup-sleep-disable-cancel pattern for this, as a dead man's switch for danngerous changes.

https://entropicthoughts.com/locking-yourself-out-with-firew...

I previously managed a firewall via scripts which would automatically revert your update in 20 seconds unless interrupted. So if you botched it and lost access, you just had to sit tight for 20 seconds.

Mikrotik has a fantastic "Safe Mode" that reverts any configuration changes that abruptly terminate your connection to the router

> cloud provider had no virtual console for some reason.

Azure still hasn't got this. It has serial and does screenshots of the console, but no access to my knowledge.

Last I checked, if you non-forcibly reboot a GCE instance via console or API and it does not shut itself down in a timely manner, there was literally no way to force it to turn off or hard-reboot so that your block storage instances get released. IIRC the last time I encountered this the process timed out eventually after some silly long wait.

Hah, I once did “netplan try” on a prototype production machine. The new config wasn’t quite right (although not catastrophic in any respect) so I told it to roll back. Bye bye new machine.

Fortunately this was an exercise and we had BMC access, so no big deal. Except that we got yet another datapoint suggesting that netplan is not a high quality piece of software.

I once accidentally rebooted the reverse proxy for all our production traffic. We got some very quiet two minutes while it came back up.

After that we installed molly-guard with a check for the number of active connections. Made it painless to reboot standby proxies and difficult to reboot active ones.

(We also instituted pairing on production proxy maintenance. I'm not a fan of pair programming but pair maintenance is great.)

I like telling junior hires about this incident because it teaches them that (a) anyone can make mistakes, (b) even serious mistakes usually aren't that dangerous, (c) you can learn a lot from mistakes with the right mindset, (d) we cannot prevent mistakes but with the right system design we can reduce their consequences.

> (We also instituted pairing on production proxy maintenance. I'm not a fan of pair programming but pair maintenance is great.)

It's a great opportunity to share knowledge and techniques and I very much recommend doing so. It's an important way to get people familiar and comfortable with what the documentation says. Or, it's less scary to failover a database or an archiving clutser while the DBA or an archive admin is in a call with you.

Also reminds me of an entirely funny culture shock for a new team member, who was on a team with a much worse work culture and mutual respect beforehand. Just 2-3 months after she joined, we had a major outage and various components and clusters needed to be checked and put back on track. For these things, we do exactly this pilot/copilot structure if changes to the system must go right.

Except, during this huge outage, two people were sick, two guys had a sick kid, one guy was on a boat on the northern sea, one guy was in Finland and it was down to 3 of the regulars and the junior. Wonderful. So we shoved her the documentation for one of the procedures and made her the copilot of her mentor and then we got to work, just calmly talking through the situation.

Until she said "Wait". And some combined 40 - 50 years of experience stopped on a dime. There was a bit of confusion of how much that word weighed in the team, but she did correctly flag an inaccuracy in procedure we had to adress, which saved a few minutes of rework.

I was using my company dev machine via Windows RDP remotely during Covid and installed Glasswire which by default blocks all traffic so I lost access. No one was there to uninstall it so I continued development in my personal machine.

Ah, but I was once a more advanced idiot. After a long console-based SSH session, I hit control-alt-del to reboot the remote computer.

I disabled control-alt-del rebooting shortly after that.

> There is no worse feeling for a programmer than waking up, walking up to the machine that was supposed to work through the night, and seeing it did absolutely nothing, stupidly waiting for hours for a response to a question that didn't even matter.

No, there's one worse feeling. Walking up to the machine that was supposed to work throughout the night, and seeing it had a surprise update that rebooted the system.

One of my favorite things about ditching Windows.

This is no longer just a Windows feature. The same thing happened to me the other day on MacOS. They recently shipped a "background" update to fix a security issue and it quite unceremoniously rebooted the machine to apply the update.

And it's why I don't use Ubuntu any more. I don't know if it still automatically updates and reboots, but neither do I care.

When did Ubuntu do that? It's been my main OS since 05 and I can't recall that ever happening?

It has never done that.

I'm reminded of this legendary HN comment: https://news.ycombinator.com/item?id=16530398

I am confused by the second guy who was curious and punched the plastic lid… it says you have to hold the button down for 30 seconds, how did that happen?

The guard itself ends up pushing the button

And it can’t be lifted back up?

Perhaps it could only be opened up with a key they did not have within 30 seconds reach - another safety mechanism that backfired in that specific scenareo.

I feel like modern tv remotes are the opposite of this principle. It is often the case that almost every single button will when pushed in some way interrupt the current program, often jumping out to a different menu or changing to a different program or something. It makes handling the remote or trying to change the volume a fraught experience.

OMG you hit my third rail when it comes to the brain-dead-designed Apple TV remote. After using one for many years I STILL press the wrong button many times every day, and in the dark, since the buttons are NOT backlit, I routinely press an unintended button. I think the user interface designer was promoted to create the Vision Pro UI/UX which is even more dreadful.

This is why I have a drawer full of antipattern design remotes (which is almost all of them really except perhaps LG Magic), and use a Logitech Harmony instead. Worst remote I've ever seen was a Sony, just a solid grid of tiny rubber squares with functionality arranged apparently at random, this sort of thing: https://i.ebayimg.com/images/g/574AAOSwevZalUBY/s-l1600.jpg. Note the grey mode switch which changes the function of a lot of the controls. It was like Sony's engineers had an internal competition to design the most unusable remote in the industry.

Fun: the “Molly” in question is Ed Krol’s daughter - he’s the guy who wrote the Whole Internet User’s Guide and Catalog.

https://en.wikipedia.org/wiki/Ed_Krol

I was almost that kid.

My dad worked for Sperry Univac. For a while he was working on a ground-support trailer for the Sergeant surface-to-surface missile. He went in to work one Saturday for some kind of a major test. For some reason, he brought my mother and I along (maybe to give my mom a break). So this four-year-old (yours truly) goes into the trailer, and sees this bright red button...

It was not the launch button. It was the emergency shutdown button, which would have cost them an hour to bring the trailer back online. Someone stopped me before I actually pushed it, but still, this did not make me popular. What I remember from that day is actually the parking lot, because I spent far more time in the parking lot than in the trailer.

The wikipedia entry lists a reference explaining the history of mollyguard, along with a pic of Ed and little Molly, but an HN comment has the relevant text excerpt. see https://news.ycombinator.com/item?id=26633835

OMG I ordered this book the second I saw your comment. It came out in 1992 and today is the first I've ever heard of it.

Oh! Then perhaps the long press required for the iPhone’s action button to trigger is a Molly guard!

Also, perhaps `rm` should be molly guarded to move things to the trash on all systems by default, and delete only if forced to by a flag.

Note: I’d have expected Molly to be a cat, because they tend to be pretty good at disrupting things in my experience.

rm has mollyguarding, that's why every invocation of rm you see on the internet is followed by -f

I think that may be a combination of (IMHO unfortunate) factors:

* Yes, on some systems rm is aliased to rm -i by default.

* Some scripts will use rm -f because normal rm returns an error if the target already doesn't exist but -f doesn't care.

* Finally, sometimes files are just ... I think it's being marked read-only that does it? I've hit this while trying to rm a git checkout; you actually do need to add -f sometimes to succeed. So if you just add -f then it'll always work.

Seeing long presses implemented for those intermittent and irreversible actions in games is something I‘ve always appreciated. I often end up making errant inputs, especially on keyboards.

A guard I often make for myself is removing/disabling the delete key on my keyboard, and setting FN+Backspace to Delete with whatever control software is involved. I often then repurpose the delete key location to F2, which is typically used to “Edit” a spreadsheet cell or file name.

Red team demur: I HATE the iPhone's long press requirement to restart from off.

To me, a button that forces me to wait for an unknown and indeterminate period of time before functioning is a FAIL.

> Also, perhaps `rm` should be molly guarded to move things to the trash on all systems by default, and delete only if forced to by a flag.

Not all systems, but some (RHEL, I think?) default alias rm='rm -i', yes

disk space is cheap these days alias to mv to trash for an extra layer of protection.

I prefer regular snapshots; I snapshot hourly. You can still miss something, but the window is small.

I once was a communications contractor for the major NJ power utility. One of their long time field techs (let’s just refer to him as Mr. T) was giving a tour of a substation that was built from the looks of it in the 50s. I have, you see, this bad habit of leaning on things… well Mr. T, without missing a beat, slid his forearm between my hip and a faded green Bakelite knob, the kind that goes in and out rather than twisting. He informed me that if I had leaned any further I would have shut off half of Newark.

Certainly sounds like he pitied the fool.

(sorry)

Sometimes a pop-up appears that I blindly accept because I happen to be typing something with spaces. Wish that button was protected somehow.

Such pop-ups should never automatically get focused. The increase of them was why I switched away from Windows many years ago, and why I like to root my Android devices. It baffles me that focus-stealing notifications cannot be turned off in most OEM Androids.

I think with power tools on Windows 11 at least it forces the privilege escalation window to pop under.

Samsung learned about molly guards the hard way - recall of millions of products after accidental fires from people/pets activating the front-panel dials.

see https://arstechnica.com/tech-policy/2024/08/samsung-recalls-...

  Luckily, there’s an easy solution recently devised that can prevent this safety hazard in homes across America, Samsung said. Customers concerned about unintentional activations can request free knob locks and covers that Samsung confirmed made it much harder to accidentally turn on the stove.

  During the meeting, the CPSC shared data showing that across 338 incidents between January 1, 2018, and May 30, 2024, stoves from “ten specific manufacturers” were involved in fires causing 31 injuries and two deaths. Additionally, the CPSC had recorded “two other fatal incidents where a range was accidentally turned on when a knob was bumped, but the manufacturer is unknown.”

  ...Companies said the CPSC data would help them “fully understand the issues” and “make sure that reasonable and foreseeable circumstances would be addressed” without impacting compliance with the Americans with Disabilities Act.

After mentioning this article to relatives, one said they had nixed buying one product because of the front dials. Then we heard from a relative in another city who bought a house due to a newborn baby - one of the additional purchases was a oven/stove/range with front panel dials.

That's wild. In Europe, we generally only have front-facing dials, but either they have a built-in push to recess function, or they require some force to turn from their off position, and for the most part heat and mode are also two separate knobs where you'd need to turn both to engage.

I never heard of any related injuries over here.

Gas stoves in particular, or at least all the ones I've used, have push-to-turn on the knobs so you can't start the gas flow by bumping them, and that includes non-auto-ignite ones where the push isn't used to trigger the igniter.

I do wish those were a thing on flat touch sensitive induction cooktops! (For all those pesky water droplets causing the cooktop to error out and turning itself off)

I get annoyed even at the thought of those things! Had to use a few while travelling. Ugh!

I personally know a guy who shut down an oil factory by pressing the molly guard button, just because the button looked interesting.

Fun random fact, Eventbrite was first a security company called Molly Guard. I spent years cleaning out the 'mg-' prefixes from the code.

best molly-guard depicited in "The Good Place": https://www.youtube.com/watch?v=etJ6RmMPGko

Lest anyone think it's just little kids that are mesmerized by the shiny red button; we were showing a potential graduate student around the compsci labs, and he walks over to an important server and simply turns it off. He could never quite explain his impulse to do so.

“Mollyguarding” sounds like a great derogation of unnecessary safety measures. Stop mollyguarding me!

I've been looking for this!

Somebody make a keyboard where every key is a molly guard, where only one will open at a time, then make a fun video about it. And credit me for this stupid idea. Even though it wasn’t my idea.

"Reverse Molly guard" is dead man's switch.

Does the disk drive or sim card slot ejector really qualify as Molly Guard?

The guard is it being a tiny hole you have to find a tool to reach into, instead of a button.

This sounded at first like a mouth guard, to stop teeth grinding.

i love the name MOLLY!

That page is copied verbatim from https://unsung.aresluna.org/molly-guard-in-reverse/ (which is linked at the top). The original page also has much better formatting.

…and Google-hosted.

Sorry about that.

Typepad, which hosted my original blog since August 24, 2004, on September 1, 2025 gave me 30 days notice that it would shut down at midnight September 30, 2025, making my roughly 40,000 (not a typo) past posts inaccessible.

I spent a frantic month trying about 10 blog hosts seeking one I, a card-carrying Technodolt, could actually use without a lot of pain.

The only one that came close was Google's Blogger.

Alas, it's horrible: janky, confusing, and always changing something I thought I'd finalized.

Oh well...

@dang Can a moderator update the link? The original is much better and we shouldn’t promote the copyposter.

Full disclosure: I posted the original and it disappeared from HN so fast it made my head spin.

Isn't it better that someone gave it a second chance, even if only by clicking a link?

No, your behavior is weird and hostile actually. Does Marcin even know that you lifted the content?

A traditional link blog would highlight a short excerpt so that the reader might be encouraged to click through to the full piece.

Yes. I just emailed him, in fact, and he responded with details, no hostility!

>your behavior is weird and hostile actually

Look in the mirror.

>A traditional link blog would highlight a short excerpt so that the reader might be encouraged to click through to the full piece.

Mine is not a "traditional link blog" nor has it ever been since its inception on August 24, 2004. You're the first person I've known to use the phrase "traditional link blog." I like it! Maybe you should start one.

You made 94 posts in the past 10 days...

What is your point? As a rule I post to HN around 10x/day, pretty much hourly.... Judging by how regularly my posts appear at the top of the HN homepage, others appear to welcome my contributions.

I just emailed a screenshot of this discussion to @dang.

I await his response.

BTW I agree about the formatting being much better. Alas, I'm limited to Google's primitive Blogger as a host so that's the best I can do.

Changed now from https://bookofjoe2.blogspot.com/2026/02/molly-guard.html above. Thanks!

Just please don't start adding molly-guards to your software. The concept only makes sense in the physical world, e.g. where the "important button", that you might never have to press, needs to be in reach all the time. In software, there are better solutions.

my favorite Debian package is Mollyguard so when you shut down a server remotely via SSH it just checks the second time to make sure you really wanted to shut down that server and not your laptop.

"Are you sure?" type guards are not suitable for actions which the user does regularly. If a user repeats this action regularly, they quickly automate the thought process (i.e. don't give it any thought anymore) and it becomes useless.

Which is why that's not what it does. It asks you to input the hostname instead, just like deleting a repo in Github does.

I know how it works. Please don't nit-pick. It's an interruption that forces the user to confirm. That's what I meant.

I discussed this also here:

https://news.ycombinator.com/item?id=46845740

It's not nitpicking. The nature of the interruption being different is material. I've lost files to automatically answering yes to rm -i y/n confirm. Typing the hostname itself is different enough to get me, at least to stop and go wait, hold on. And snap me out of doing the wrong one. Especially an SSH gateway machine.

Reminds me of this Matt Levine

>> At 08:56 a ‘Trade Limit Warning’ pop-up alert appeared within PTE. This presented the trader with 711 warning messages, consisting of hard block and soft block messages, listed in a single alert where only the first 18 lines of alerts were immediately visible unless the person who received the alert scrolled down. The trader did not appreciate their inputting error and overrode all of the soft warnings in the pop-up.

> You get 711 alerts, you only see 18 of them, you are like “ehh 18 alerts is pretty much the normal number,” you override them all without reading.

I agree. Fortunately, molly-guard the software can be configured with automated checks to allow safe actions (e.g. shutting down servers that don't receive significant traffic) without pestering the user.

This means a properly configured mollly-guard is invisible for routine actions but kicks in only when a genuine mistake is suspected because the operation would cause some sort of meaningful loss. That way, users aren't trained to ignore it.

> can be configured with automated checks to allow safe actions (e.g. shutting down servers that don't receive significant traffic)

That's clever. This is what I meant when I wrote, that software allows for better solutions.

Reply on news.ycombinator.com