Old-school computing: when your lab PC is ancient

https://www.nature.com/articles/d41586-021-01431-y

by sohkamyung • 5 years ago

123 154 5 years ago

linksnapzz• 5 years ago

Oh, the days of 30-yr old PDP-8s running CNC milling machine s in the physics dept. machine shop....

inetsee• 5 years ago

I worked as a lab assistant in the Pychology department at the University of California at San Diego from 1974 to 1976. The lab I worked in had a PDP-8 minicomputer. While I was there they added a 32K word hard drive for $5000, and upgraded the core memory from 6K bytes to 12K bytes for another $5000. I sometimes think about what I could get nowadays with $10,000.

spfzero• 5 years ago

And that’s 10,000 1976 dollars. Today that would be at least $25,000, or more.

pmlnr4• 5 years ago

a PDP-8 is not 30 years old... it was released in 1965. The last batch was in '74, according to wikipedia. That's ~50 years old. Remember, 2000 was 20 years ago.

bitwize• 5 years ago

Grandparent post was reminiscing, ostensibly to a time about... er, 20 years ago.

analog312• 5 years ago

Thank heaven for RS232 and documented ASCII command sets.

Having to replace working equipment because the software that you paid for can't be transferred to a new computer is a nightmare Often, the old equipment works just fine because the laws of physics haven't changed.

But equally often, by the time it's that old, it's being used for a tiny subset of its original, perhaps for exactly one repetitive procedure. In those cases, it's often easy to whip up a little Python program that talks to the RS232 port via a USB-RS232 converter.

Will the Python program last any longer? Maybe not, but at least its source code is in text format.

j454• 5 years ago

Tibbo makes some neat products to do rs232 over Ethernet

iancmceachern• 5 years ago

Was looking for something like this for a project, they are neat, thanks!

simlan• 5 years ago

Yes. I always rejoice when i find a scan of an old manual somewhere and the last pages consist of that delightfull command list

anthk2• 5 years ago

Will the Python program last any longer? Maybe not, but at least its source code is in text format.

TCL and Expect would be much better for that than Python.

hansor• 5 years ago

I my previous job we used "Moxa Serial to Ethernet Converter" [read-only in one way] to connect with almost 100 machines in various laboratories around the country to get spectography(and other "BIG OIL" stuff) data into my distributed SQLITE and MS-SQL databases back in HQ.

Worked really well. One of the oldest machine was some 286 connected to some freaky gasoline density aparatus :)

In many cases you can't simply upgrade software, as software for such machines does not support anything else - and you can't really reverse engineer drivers for 50 different machines in sane amount of time.

Most of really expensive machines provide some kind of output over serial interfaces - it seems to be industry standard, but I forgot the name for it.

the_only_law4• 5 years ago

Surprised a lot of this stuff is RS-232. I have a collection of old engineering/telecom test equipment I used for a project and while some of it did offer RS-232 connection, GPIB is probably a bit more common.

analog312• 5 years ago

Depends on the industry. Chemistry lab equipment like spectrometers more frequently used RS232. But someone does make an inexpensive USB-to-GPIB adapter that basically enumerates as a COM port, making programming pretty easy.

asdff2• 5 years ago

I used to work at an institution where we had this fluorescently activated cell sorter machine that was hooked up to a computer running windows XP and the proprietary software you of course cannot upgrade (as outlined in the article). Since it was so old, IT airgapped it. No USB sticks even. Data left that machine via freshly burned CD rom. It was no 80s machine at least.

eric__cartman3• 5 years ago

I really hope they saved a full disk image to restore in case the hard drive failed.

asdff2• 5 years ago

You weren't supposed to leave your data clutter on the machine after you burned it onto your disc, but people certainly got lazy with that. If it did crash one day, IT would just have to find another pc running XP to install the proprietary software on. Plenty of ancient equipment floating around random closets at that place. Benefits of a physical installer disc.

secabeen• 5 years ago

As a research IT professional, this is our everyday. Not sure why it justifies an article in Nature, but cool to see some exposure.

eru• 5 years ago

I guess it's good to have some mundane article in Nature.

Because otherwise Nature mostly attracts high profile papers, that means that articles in Nature are more likely to be wrong than articles in other journals. (Interesting findings are more likely to be wrong.)

Fomite• 5 years ago

I think the idea is to highlight this problem - especially as more and more labs have complex machinery with a PC as an essential component of it.

And that this task is often not the responsibility of a research IT professional, but a puzzled grad student.

ironmagma• 5 years ago

A lot of people will not believe something is real unless there’s an article that can be cited about it. So I guess it helps for when someone demands a source.

fsh6• 5 years ago

Only the back half of nature magazine contains scientific papers. The front half is articles like this one that are meant for a general audience. That part part is surprisingly readable and often quite interesting.

op00to2• 5 years ago

It’s a huge problem, and only getting worse. It contributes to the problem of non-repeatability in science. Also the inability to fix stuff by throwing money at it was one of the reasons I left research computing.

unethical_ban• 5 years ago

I'm not sure why it doesn't justify an article in Nature.

vnxli• 5 years ago

I think most of the world is like this. New equipment means big CapEx asks. A lot of manufacturing runs on an old XP machine held together by ebay parts and duct tape

userbinator• 5 years ago

In some industries the machines can be around a century old if not more, and were converted to computer control only "recently" (as in decades ago). From that perspective, the rapid change in computing is the anomaly.

eric__cartman3• 5 years ago

That's the beauty of open source software and standard protocols like RS232. No matter how long ago the software has been developed you can port it to newer (and more mantainable) machines.

adrianN2• 5 years ago

All the big machines with computers in them run really old hardware. Airplanes, trains, etc have lifetimes measured in decades. The manufacturers stockpile parts, because these old chips are no longer for sale.

walrus015• 5 years ago

One piece of advice I would offer to people using really old x86 hardware for lab or other purposes, which has hard-to-replace software on it. Find a way to use clonezilla to take an image of the whole (fat16?! fat32?) filesystem, and also of the disk as a whole including partitions. Store it somewhere safe in .tar.xz format. Additionally take images of whatever original install media you can find.

I'm talking about stuff like where you find a system running a CNC milling machine from 1992 with a 60MB IDE hard drive in it.

You might need to pull the hard drive and connect it to a somewhat newer system in order to boot from clonezilla and take the image. Something from the early 2000s with an ATA/33 controller on it should be backwards compatible to almost all of the oldest IDE hard drives. Can be quicker and easier than trying to run clonezilla on the native hardware (or on systems which have no idea of how to boot from a CD-ROM or USB).

If you have something a lot older than that, it can get more complicated. Never personally needed to take a disk image of something with a MFM/RLL ISA HDD controller card driving an old 5.25" 10MB/20MB type drive, but it's probably possible. I bet you could get the controller card working in a 1999 vintage Pentium 3 in the ISA slot with enough hackery.

phendrenad2• 5 years ago

Yeah definitely. I swear, those old hard drives have a way of KNOWING when you're about to back them up and failing, just to spite you.

eric__cartman3• 5 years ago

That's because you need to sneak behind them with a Hiren's Boot CD before the hard drive has time to kill itself.

geocrasher3• 5 years ago

I once had a hard drive fail this way. It was a 60GB SATA disk from around 2006. I powered down the machine to drop in a couple of new disks and gave the machine a much needed dusting.

When it restarted it had the click of death. The freezer trick worked long enough for me to grab the few kb of data I really needed from that disk.

I wrote about it here: https://miscdotgeek.com/adventures-in-hard-drives/

bserge• 5 years ago

Most likely because the heads try to read a few back blocks that weren't accessed for years.

You can do block checks regularly, which takes a looong time on high capacity drives (or is impossible on newer ones due to firmware restrictions).

dosman332• 5 years ago

If you are powering off a very old system that spends most of its life powered on, don't let the hard drives or power supply cool down. Do your best to keep your work as brief as possible and get the thing turned back on before it assumes room temperature. These components tend to run the hottest, you can have cold solder joints contract and stop making contact when they cool down.

Also check any socketed chips and make sure they are fully seated. Chip creep happens when frequently thermal-cycled boards cause the IC's to slowly lift themselves out of their own sockets.

j454• 5 years ago

Ghost and VMware converter are good tools for this as well

walrus015• 5 years ago

As I recall the origin of Clonezilla was as a GPL/open source licensed community built alternative to Ghost, but to fulfill exactly the same functionality.

LeoPanthera• 5 years ago

If recovery is critical, use lzip and not xz. The compression ratio is the same but xz archives are particularly vulnerable to corruption.

mangamadaiyan2• 5 years ago

Out of curiosity, what in the file format makes it particularly vulnerable to corruption?

adrianN2• 5 years ago

If recovery is critical, store some parity bits too. par2 or something like that.

squarefoot2• 5 years ago

I wouldn't use compression at all wherever possible. Those old drives are likely at least two orders of magnitude smaller than modern ones, so storing everything uncompressed, or even the entire disk image shouldn't be problematic. Plain uncompressed files would also be easier to recover in the unlikely, but possible, event of failure of the media used to salvage them to.

dmos62• 5 years ago

I think hard drive images are rarely compressed, because it's just so slow.

Chris2048• 5 years ago

Also, it can be a nice trick to take a fs-structure dump: list of all files on disk, with associated sizes, permissions, and md5 hash.

It should be much smaller (and faster) to take such a snapshot, but priceless if you ever suffer corruption for e.g identifying & verifying inodes, restoring permissions etc.

Also useful for identifying which files are vanilla (same as installed from package) so can be restored by other means (finding original package). For that reason, a dump of installed packages w/ version is also useful.

cookiengineer• 5 years ago

This is literally the reason why I keep my ancient panasonic CF and 1st gen latitudes and thinkpads around.

The laptops have the huge advantage that they have a serial port, which allows to rescue pretty much everything given that you implement an adapter software for it.

I wish modern hacker laptops had an easy non-proprietary way to get a serial port.

The only alternative we have is using an Arduino based solution...but that's like a downgrade compared to 20 years ago.

midasuni• 5 years ago

USB serial adapters will give you a standard usb port at /dev/ttyUSBn, usable via minicom etc.

kogepathic• 5 years ago

I wish modern hacker laptops had an easy non-proprietary way to get a serial port.

Dell docking stations like the PR02X provide serial/parallel ports directly from the laptop, and Dell Latitude laptops with the E-Port docking connector were made until ~2016.

Both the PR02X and Latitude laptops with an E-Port are readily available on eBay for reasonable prices.

I've successfully used an E6220 with the PR02X to bitbang out the parallel port for a project.

the_only_law4• 5 years ago

I have an old toughbook for for this reason as well. I tried looking into some modern ultra ruggeds but damn they’re expensive.

jlgaddis2• 5 years ago

About a week ago I wrote in a comment here that "Some of us (masochists) actually like RS-232!". It's nice to be reminded that I'm not the only one!

My primary laptop is an (8-year-old!) ThinkPad W530 that I can't bring myself to part with (it's a beast and still outperforms many of today's more "modern" laptops).

It doesn't have an actual serial port built-in but it does have an (34mm) ExpressCard slot. Luckily, I managed to find a card with an actual RS-232 serial port that connects to the host using PCIe, NOT USB.

I'm going to be quite sad once the W530 gives up the magic smoke. I have stockpiled spare hardware in preparation of that day, however.

zrm2• 5 years ago

You might need to pull the hard drive and connect it to a somewhat newer system in order to boot from clonezilla and take the image. Something from the early 2000s with an ATA/33 controller on it should be backwards compatible to almost all of the oldest IDE hard drives. Can be quicker and easier than trying to run clonezilla on the native hardware (or on systems which have no idea of how to boot from a CD-ROM or USB).

Also note that Parallel ATA host bus adapters exist and then you don't even need an old PC, just a ~$20 adapter for a modern one.

walrus015• 5 years ago

although some modern PC BIOSes have problems with manually entering the heads/cylinders as you might find on the label of an IDE hard drive from 1992. additionally I have found that modern udma/133 parallel ata drive controllers have less than perfect compatibility with some very old IDE stuff. while a 'vintage' motherboard like something with an intel 440fx or 440bx chipset from the very late 1990s should work.

pmlnr4• 5 years ago

https://en.wikipedia.org/wiki/Ddrescue can come in handy if the situation is bad.

grishka3• 5 years ago

If you have something a lot older than that, it can get more complicated.

Could probably dump it out through the serial port.

nix23• 5 years ago

.tar.xz

No, use tar.lz

http://lzip.nongnu.org/

http://lzip.nongnu.org/safety_of_the_lzip_format.html

http://lzip.nongnu.org/xz_inadequate.html

vmchale• 5 years ago

Store it somewhere safe in .tar.xz format.

Or lzip! Has better recoverability properties: lzip.org

jlgaddis2• 5 years ago

Find a way to use clonezilla to take an image of the whole (fat16?! fat32?) filesystem, and also of the disk as a whole including partitions.

And perhaps invest in a few CompactFlash <-> IDE adapters, too.

LocalH• 5 years ago

Do be careful about powering off HDDs that have been possibly spinning for 20+ years. They might not spin back up.

walrus015• 5 years ago

Yes absolutely. If it's running and you can get access to I/O on it to copy data off via serial or something (let's say it's running MS-DOS 3.3, 5.0 or 6.2), use something like laplink to transfer as much off as possible before taking it down for intrusive maintenance.

https://www.google.com/search?channel=fs&client=ubuntu&q=lap...

every• 5 years ago

I once worked for a large, urban Episcopal church. Since it was non-profit, parishioners could donate their old computers for a tax write-off. Amongst my many duties, I converted these hulks into network terminals. They ran Novell (DR) DOS on a Novell 3.12 network. I used third world cheapo compatible NICs since my annual computer budget was in the few hundreds of dollars. And as long as the patch cord wiring system held up the thing actually worked...

stordoff• 5 years ago

I'm not sure if it was necessary or they just hadn't needed to change it, but one of the undergrad physics practicals at Cambridge used a Windows 9x machine (I think it was 98, at least as of 2008). Seeing "It is now safe to turn off your computer" again at the end of the session was oddly comforting.

fsh6• 5 years ago

I'm always a bit worried about using a Windows 10 machine in the lab. You never really know when it will decide to update and randomly reboot (or show a fullscreen ad for Edge). Unfortunately with the EOL of Windows 7 there aren't many alternatives. Lots of lab hardware (especially from small manufacturers) is still Windows only. That is a problem you didn't have with your 386 machine running DOS.

yjftsjthsd-h• 5 years ago

Isn't that precisely what LTSC is meant for?

asciimov2• 5 years ago

If you can get a license for it and if the software will run on it.

noobermin2• 5 years ago

I have been trying to get my experimental group to move off their computers off windows but I unfortunately don't have that sway, plus they're wedded to labview. That's the one value of linux even with its warts.

laserphysicist_3• 5 years ago

We managed to get away from proprietary software for our simulations needs and moved to oss that runs on both Windows and Linux. But we have not been able to get rid of Labview & Windows in the lab. I'll be happy to hear about your experience here...

pletnes2• 5 years ago

Even on windows xp/7, I know people running hours-to-days long simulations routinely unplugging their ethernet so the IT department / windows daemons / tooth fairy won’t randomly reboot it and make a project late.

laserphysicist_3• 5 years ago

If only we could do that! But with ldap accounts, nfs-mounted files and Labview licence server, everything stops working whenever the network is down/unreachable...

bellyfullofbac• 5 years ago

There are surely certificates with expiration dates on these systems, so when the internal clock goes beyond that date, some stuff will stop working. I guess then someone would write a date mapping tool (calendars repeat themselves every 28 years, with matching weekdays and leap years, although in 2100 this rule will break).

miked85• 5 years ago

(or show a fullscreen ad for Edge)

I am not a Windows user, so maybe a dumb question, but is this actually a thing?

codingdave• 5 years ago

I am a windows user, and have never seen one. I'm not saying they don't exist, but it sounds more like something added on by whomever manufactured the PC, not a core part of the OS.

wizzwizz4• 5 years ago

I've never seen one. Centre-of-the-screen and taking up about 50% of the area, yes, but never a maximised or full-screen window.

fsh6• 5 years ago

I have seen this screen after an update on my laptop: https://www.windowslatest.com/2020/11/15/windows-10-is-now-n...

grishka3• 5 years ago

Connecting your critical equipment to the internet is never a good idea. And if you don't connect it to the internet, the support status of the OS it runs shouldn't really matter much.

fsh6• 5 years ago

There is a very large region between "has to be completely air-gapped" and "better should not reboot randomly".

SAI_Peregrinus• 5 years ago

It's possible to "break" Windows Update reversibly (I discovered this by accident on a computer with a large HDD and small SSD a while back). Change the location of the default `ProgramFilesDir` and `ProgramFilesDir (x86)` keys in `HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion` to not be `C:\Program Files` and `C:\Program Files (x86)` respectively. EG move them to a different drive. Don't move the various programs inside. IME, everything works, except that Windows Update will check only the new location for some Microsoft programs, not fall back to the default, and fail with an unintelligible error code.

ehnto2• 5 years ago

One of my fondest memories was accompanying my then partner to another city, so she could use their particle accelerator for her research. I didn't get to go anywhere near the accelerator and so I just wandered the city, had dinner and whatnot.

It was mostly operated from a command line and she was run through how to operate the equipment and do her tests, then left pretty much to her own devices with the accelerator.

At a bit after midnight, I get a couple of text messages, her firings all went to plan, but she can't figure out the command line in order to archive and copy the data to her personal drive. Cue a 1am phone call trying to figure out what the operating system was and what commands were at our disposal. Thankfully it was unix based and everything went smoothly, but it was fun and I was glad to feel useful on the trip!

101008• 5 years ago

Nice story, I got a bit of panicked after I read about that call. I can iamgine her fear of losing all the hard work becuase she was not able to export it.

ehnto2• 5 years ago

I was pretty anxious too. We ended up using tar, and it felt exactly like that XKCD comic on the topic:

https://xkcd.com/1168/

geocrasher3• 5 years ago

I did work about a decade ago for a company that made trophies. The kind you win and sit on a shelf. Their engraving machine was running on an original IBM XT or AT, with two 5.25" floppy drives and a couple of old school floppy disk containers on the desk which contained all their fonts, etc. I warned them that the machine would fail eventually and they should replace it soon. "But it still works."

Indeed. It still worked.

hellbannedguy• 5 years ago

They used to make computers that lasted.

rytis• 5 years ago

The computer is still going to work, no problem. Floppies on the other hand are bit riskier bet.

coldpie• 5 years ago

They still do. The problem is the software.

nikanj• 5 years ago

In 30 years we'll have plenty of computers from 2020 that are still used. Survivor bias is strong, we just don't see all the computers from the 90s that already gave up the ghost

geocrasher3• 5 years ago

I did a fair bit of work around 2010 for a fellow who ran a EEG machine at a neurology clinic. It ran Windows 98, and was connected to the Internet. I cleaned that thing of malware more than once. Replacement was prohibitively expensive unfortunately, as it require replacing the entire unit, not just the PC portion of it.

noobermin2• 5 years ago

Reading this then remember comments on that article about the AMD's new thing taking advantage of TSMC's 3D fabrication where a datacenter worker said they'd use CPUs "up to seven years" insinuating that's a upper time limit that would be undercut as the chips "go out of date" sooner.

Scientific computing (computational science) is a pain in many ways but the aversion to churn and honestly waste the rest of the cs/it world takes as natural is one small valuable nugget. I wish devs in general had a longer outlook when the make things in general.

incanus772• 5 years ago

I occasionally troll Craigslist locally for old hardware for my collection and for fun. I recently got a top of the line — for the late 90s — Graphtec 24” vinyl cutter/plotter, complete with PowerMac 9500 with CRT and Jaz drive all for 40 bucks. It works great, has Mac OS 9 and Illustrator and everything else that I need for the setup. The owner’s dad ran a sign business until he passed away, then the son used it for hobbies for years. I’d like to modernize it a bit and shrink the Mac out of the equation, but the system just works for now.

amatecha• 5 years ago

Nice, that's a pretty good deal! I just dug up an old Jaz disk myself -- just have to figure out which of my machines (if any) has a drive.. Ahh, the good ol' days ! haha

BBC-vs-neolibs• 5 years ago

Possibly you could run the full stack with Qemu or https://sheepshaver.cebix.net/

Might even work fast enough on a Raspberry Pi 4.

incanus772• 5 years ago

The comms protocol itself is sending to 25-pin serial at 9600bps, so it’s pretty simple to rig up another way. The stuff sent is just monochrome vector art. I was thinking of either finding or making an open source (maybe Python) lib to talk to the protocol and then yeah, probably running it on a Pi as a service I could send right to from my computer. The plotter is about waist-high, so it’s pretty large. I figured I could velcro a Pi to the underside pretty easily and serve via wifi.

TacticalCoder• 5 years ago

Forney at one time had a stack of ten of her 1987 Toshiba laptops that she’d bought on eBay for $10 to $12 each. “We’re down to our last one,” she says.

I don't remember all the details but if my memory ain't failing me there was something similar with the software used to interface with the iconic McLaren F1 car. Only about 60 of those cars have been produced (at least for the road, maybe more for competition) and the mechanics servicing these cars had to stockpile old laptops because the proprietary software/interface had never been ported to modern technology. Eventually a few years ago McLaren assigned developers to the task and a new version was made. I think they did it because they were running low on the pile of old rusty dusty laptops ; )

fock2• 5 years ago

iirc this even happened to BMW, where their config software ran on notebooks with Siemens(!) UNIX up to mid 2005 or so.

mrpippy• 5 years ago

Oh wow, I didn’t know this before: BMW’s DIS (Diagnosis and Information System) runs on SCO OpenServer 5. And this was something used in the field by mechanics!

pmlnr4• 5 years ago

“If that board dies we’ll have to go onto eBay”

from this article, whereas

"I have a prospective customer supporting a U.S. missile defense system that is buying parts on eBay,"

from

https://www.pcworld.com/article/249951/if-it-aint-broke-dont...

If it ain't broke, don't fix it.

rob74• 5 years ago

To work with current Windows versions, the team would have to replace the whole microscope. The marginal potential gains aren’t yet worth the US$400,000 expense, Cherr reasons.

Unfortunately there is not much information given on this particular example, but that definitely sounds like throwing the baby out with the bathwater? If the Windows XP computers are just interfacing with the microscope, surely it would be cheaper to just reimplement this interface (of course, the alternative of leaving everything as it is is even cheaper, until you get a ransomware infestation)? Or are they talking about Windows XP Embedded directly controlling the microscope?

freeone3000• 5 years ago

They're talking about Windows XP, client, directly controlling the microscope -- usually through a 16-bit driver. Vista redid the driver stack and broke compatibility with these. The driver is provided by the device manufacturer, who has little interest in providing new software for old hardware.

op00to2• 5 years ago

How do you know your reimplementation isn’t subtly changing data? All research must be reproducible, and that includes the instrument and software chain as well.

asciimov2• 5 years ago

I've worked with similar microscopes. One of them, worked perfectly fine for the tasks it was assigned to, but the software would only work up to Windows XP.

To make matters worse, the whole system was hostile to any changes or updates. We didn't have an installer for the program but disk images to restore the system with the controller software already installed. The computer also had a hardware key connected to validate its licenses. Finally due to some validation checks the software performs, the controller software can't be placed on a virtual machine.

One way you can deal with this is by getting the longest service contract you can afford when you buy the equipment. We had a slightly newer microscope with a service contract that was still good when the XP to Win7 transition happened. It took some prodding of the vendor, but they eventually came down and replaced the controller PC free of charge. Eventually those service contracts run out, and you are left with a perfectly good microscope, and an aging computer.

azalemeth• 5 years ago

One experience I was lucky to have as a graduate student was bumping into a set of four old SGI Indy O2s running SGI Irix and driving an old (~£5m) MRI scanner. As a child of the 80s, SGI Irix was the missing OS that I hadn't actually used: Mac-like in GUI but with solid unix underpinnings. It was -- in my mind, at least -- the "Mac OS 9" version of OS X.†

Actually driving the things was a slightly different experience. The software was horrible to program in, old, klunky, but feature rich and very hackable. It made me question my gnu/BSD knowledge every time I typed, e.g. 'ps auxw' and got a usage statement; `vi` was ancient and definitely not vim running in a compatibility mode; and the `del` key issued kill (or maybe SIGHUP -- can't remember) signals. I learnt a lot.

That machine was finally replaced when a professor nearby moved on, his former department didn't want his old machine, and I commandeered the department of physics's official (and excellent) Man and Van to help replace it with another microimaging system that I built using a mix of old hardware and open-source software [1], for which I am lucky enough to have lots of electronic schematics to go with and the relevant firmware. It's still going now, several years later. My old PI is still trying to get the money to replace it with a modern Bruker instrument -- which will be far less hackable, far more locked down, and come with a hardware DRM dongle; but doesn't require a postdoc (like me!) to (re)build and maintain.

† Yes, I realise that was NeXT, but I've still never actually seen a NeXT machine in the flesh. I'd love to!

[1] https://openvnmrj.org/

spongeb00b• 5 years ago

My high school physics dept were still using a then 20 year old BBC Micro to run experiments which only worked with a specific piece of ultrasonic experiment gear.

I wonder if it still gets wheeled out on that cart to this day...

RalfWausE2• 5 years ago

At the company in which i am the head of IT, we are - right at this moment when i type this - running a production batch of stuff we produce on a filling machine controlled via an C64 board. And it "just works", and as long as you can buy a working C64 for under 100 Euros i would see absolutely no need to change anything.

And frankly: I like it ;-)

mattowen_uk• 5 years ago

I would explore whether the Ultimate-64 FPGA-based C64 clone board would work for you, as that would give you long term confidence on keeping the platform running.

RalfWausE2• 5 years ago

Thank you for the hint, i will look into this... i think my boss would be happy if he knows that the > 100.000 € production line is not dependent on some obscure tech his "it guy" is keeping somehow alive ;-)

_trampeltier4• 5 years ago

We also had to buy a new device to measure something on work (>100000 CHF) just because an old PC died with Win3.1. PCs in industry in the long term are just a pain in the ...

detaro3• 5 years ago

weird custom hardware that couldn't be replaced? or needed to be "official" and couldn't just be changed?

_trampeltier4• 5 years ago

A lot of old things had custom hardware.

dsr_• 5 years ago

I see a common problem with all of these examples.

Equipment without well-documented interface protocols -- or better yet, open protocols.

It was a lowly printer that ended up infuriating RMS and establishing the FSF and the GPL.

the_only_law4• 5 years ago

some computers still run on Microsoft Windows XP (released in 2001), because of the need to maintain compatibility with a scanning laser confocal microscope and other imaging equipment

Damn here I was hoping “ancient” meant some workstation from the 70s/80s. Should have known better tbh.

dosman332• 5 years ago

You're dating yourself. XP from 2001 is 20 years old now. I get it, I cringe when I realize this too.

tantalor• 5 years ago

On the software side, is emulation/virtualization a good solution here?

bombcar2• 5 years ago

It can help but often these computers have a relatively custom control card built in - think ISA board designed to connect to some piece of machinery.

Which is why things like this exist: https://www.assured-systems.com/us/product/atx-intel-g41-cor...

And it gets more complicated when you have non-x86 hardware - you may be perfectly able to emulate the software but have no way to connect to the controller board (think Apple II controlling a machine).

nahuel0x• 5 years ago

There is a market niche here for a computer with every lab interface under the sun (ISA, RS232, RS485, BNC Ethernet, etc) capable of running/backuping virtualized legacy OS's and with a clear upgrade line and good support. There are "industrial PC's" with legacy interfaces, but the software side is not resolved.

itomato• 5 years ago

I worked at a company that maintained racks of DOS boxes to run a some custom code. The contractor that wrote the app also maintained the boxes.

My role as Sys Admin meant I was ready to decommission all four racks in favor of 4U worth of hypervisors.

He dug in, fought dirty and the whole company went under.

qubex• 5 years ago

I’m going to draw from the article and from my personal experience (somewhat related viewpoint of computers in small companies’ chemistry labs) and point out the leading cause of this obsolete technology persisting: cost (or rather, price).

Mainly we’re dealing with deliberate decisions not to upgrade exclusively based on price. This is quite disturbing (as an economist). It’s probable that by keeping the prices high on software the companies reap benefits from those clients that are willing to pay (presumably for updated specifications and algorithms but mainly for compatibility with newer systems), they’re also relegating users by not providing a more convenient and reasonably priced upgrade path that covers merely the compatibility aspect.

euler_angles• 5 years ago

I used to (2012 - 2014) work on that Sounding Rocket program the article talks about. I never saw a monochrome screen, but the computers handling the telemetry monitoring during launches were old Dolch lunchbox style cases holding a single board PC with a really cruddy LCD panel. The computers were either 386s or 486s (depending on which one you got) running DOS 6 and a program written for telemetry monitoring in 1989 called 'tdp502.exe'. This was written in Borland Turbo C++ and was rather flaky.

Coming off a program with 2000s era equipment for telemetry and monitoring, going to this was a big shock and the pain of using such antiquated equipment made everything very difficult.

______-• 5 years ago

I always found it odd that SSDs die out within 5-10 years if they're not accessed frequently (i.e they're not suitable for long term archival storage) yet oldskool computers' storage of the 80's seem to last forever. It almost makes me want to buy some of this old tech since the storage longevity is markedly better.

vikingerik• 5 years ago

Well, survivorship bias. Nobody is talking about their 80's computers that died by the 90's. And nobody can yet talk about tech from the 2010s that has lasted 40 years.

That said, yes older storage methods may have better longevity. Compare an SSD to game console ROM cartridges that last virtually forever. What fails in both older and modern storage hardware isn't really the per-bit magnetic storage, it's the environment necessary to access them, the power supply and spinning motor and moving head in a hard drive.

zwieback• 5 years ago

Have you bought a hp ink cartridge lately? Chances are it was inspected by one of my machine vision system running on an old XP PC using old PCI cards and analog cameras. We're slow to upgrade these systems once they are running, especially if they've been transferred to outsource suppliers somewhere in Asia.

beh9540• 5 years ago

This reminds me a lot of work I used to do in the controls world. We had a large amount of equipment all with proprietary networking to embedded systems that were basically Intel 386 systems running DOS. The big issue we had was they had spinning hard drives in them, which would fail, and replacing them meant we needed to replace all the nodes on the network, often hundreds of pieces of equipment. To make matters worse, the company was long out of business.

My boss came up with the idea of replacing the drives with industrial flash drives, so as the drives failed we replaced them. We had some issues getting drives that were small enough, as there was an issue with the BIOS and drives that were too big, but we were able to keep the systems running so we could eventually replace the whole networks while equipment was getting upgraded.

We had an even older system that was an entirely custom mainframe computer with 10" hard drives and modem banks that we all stayed far away from. It wasn't as critical, so it was even slower on replacement.

Old-school computing: when your lab PC is ancient

Related Stories