Hacker News Reader

Comment by Workaccount2 | Hacker News Reader

My company provides new compute modules for another major global city's infrastructure that still relies on Intel CPU's from the early 80's. That is, we are building them brand new boards populated with chips that are over 40 years old.

They haven't shown any interest in updating the system. It works, they can get service, and get "new" replacements for things that go bad.

What they might not know though is that there is basically just one engineer we have (and probably the only one on Earth) who knows how to work on these things. He's getting old, and obviously none of the younger engineers really have an interest in learning ancient forgotten systems.

View on HN

LOL:) this story seems to resurface every couple of years. My favorite part of the article:

"Jeffrey Tumlin: "It's a question of risk. The system is currently working just fine but we know that with each increasing year risk of data degradation on the floppy disks increases and that at some point there will be a catastrophic failure."

This seems to imply they have been using the exact same disk for the past 20 years (absurd), they have absolutely no idea what is written on the disk and how it can be safely backed up or restored. This would be a problem regardless of the medium used.

Although I hold the line at using paper tape there is nothing wrong with using floppies other than it seems antiquated. It certainly is reliable and cheap. Maybe the only thing that needs replacing is the people running the Muni.

You make a good point. However, even with the disks backed up to a better storage medium, detecting data loss on the floppy and getting a replacement written and deployed to the correct location might take a while. During that time the system may face unacceptable transit delays.

Wait, what? If this thing boots from a floppy, and that floppy has a checksum on it, and the boot sequence involves loading the checksum and the rest of the floppy contents into RAM and computing the checksum, corruption detection would be near instantaneous and the remediation would be as quick as it takes to eject the corrupted floppy, grab another floppy copy from the bin, plug it in, and flip the on switch. So the delay would be (roughly) zero.

Let's say it's not just booting from it, but constantly reading.

Let's say no one implemented the checksum. Edit: I forgot that floppies have a crc check, my bad.

Let's say the machine that can write the new disk is physically far away from the machines that read them. (Or the guy with a stack of new disks -- which are harder to find every year -- his office is some distance from the control machine).

I guess I assumed it was only reading from the floppy, not also writing to it.. that doesn't seem like it would last very long at all without encountering errors (at least based on my memories of reusing floppies until they were completely worn out)

Every sizable manufacturer of floppy disks has exited the market. There are no more (major) suppliers for the tech. They’ve likely been depending on a dwindling supply of functional disks; if at some point they find themselves without enough working disks to operate the system, they will indeed be screwed.

There’s also the chance that they take a disk that’s on the verge of failure, plug it into the system, and some corrupted commands get loaded into the system. That could easily result in a “catastrophic failure”.

Floppy disks are not reliable or cheap. They physically degrade over time, and at this point are nowhere near cheap for “new” disks.

One of my older jobs relied on ancient hardware, and the life of the company was measured by parts in the warehouse and vendors who still cared. From their perspective they'd have a modern system as long as your company delivers.

> hey haven't shown any interest in updating the system.

So expensive to update that there's a calculated end-of-life to the system. They'd love to know about your engineer situation. That'd trigger plans put in place a while ago.

Your company could do a last big batch for the city and send the old guy out with a nice bonus.

Nah getting you fee engineer interested isn't that hard, there's always a fun challenge in mastering something. You just have to give them time to figure it out.

At my workplace, we still ship devices with AMD 8086 processors to the military. And AMD still makes us 8086 processors on special order.

We got a few guys under 40 that work on that project once in a blue moon when a change is required.

Are you hiring young engineers who want to learn the old stuff?

Where are you going to find a young engineer who wants to build a dead-end career? I'm sure that if you pay them enough, someone will show up, but most people would rather build skills that will continue to be useful rather than skills that will be useless once an inevitable, sorely needed refresh happens.

I think you overestimate how much working on obsolete tech is a dead end career. Skills are still transferable. Someone who can sling 8086 assembly can figure out your dumb web app.

I’d do it. I’d like to be the expert that can help them choose or maybe build the replacement.

The replacement will use modern parts and tools, administered by someone who knows modern systems.

There is no use in becoming an expert in an ancient system ridden with proprietary and obsolete parts, just so you can help one city maintain their ancient system. It's even worse than wanting to become a professional fax machine guru.

There is no use for so many things… For example there is no use in assuming you know what other people want to do with their time, because you just don’t know. Some people like knowing unique things, or just enjoy learning, or want to be a small hero for a small group of people, or are just curious.

Keeping a transit system that is relied on by millions of people running seems like a very useful endeavor.

The problem, for an engineer considering this path, is that the gig ends before you want your career to end.

Reply on news.ycombinator.com