News MS-DOS and Windows 3.11 still run train dashboards at German railway — company listed admin job for 30-year-old operating system

[rluker5]

Whoops!

In 2002, my job gave me a P4 (Northwood) 2.53 GHz with 512 MiB RAM as a development machine. The GPU was an AMD 9700 Pro and the monitor was the first LCD I'd used.


Eh, I'll say Windows NT 3.5 was good. That's the earliest version of it I've used.

Yeah, you're right. If you click the right part of the logo on Apacs+ it gives you system specs and it pops up 5xx,xxxKB for the ram. I was just thinking of the number displayed.

 

[aldaia]

That is nothing. There are many critical systems ( I would say much more critical than German railway system) running on much older computers.

Some GE nuclear plants still use PDP-11 computers and they plan to do so until 2050s. PDP-11 is a 16 bit computer released in 1970. DEC, the company that produced them, ceased to exist in 1998.

 

[DenisBBergeron]

That's something I did a long time ago, but I wish for them, that their MS-DOS and Windows 3.1 run under a VM on a linux machine.

 

[Darkoverlordofdata]

Outside of the tech industry, the board of directors don’t always care about the tech considerations. I worked for a major insurance provider in the ‘90’s that ran the business on obsolete IBM equipment. Thay had a warehouse full of old used IBM, and hired old IBM engineers to keep it hot swap ready.
BAs said the newer software would’ve required changes to the business structure, and the board wouldn’t approve an upgrade. IT loved it - job security all around!

 

[bit_user]

That is nothing. There are many critical systems ( I would say much more critical than German railway system) running on much older computers.

Some GE nuclear plants still use PDP-11 computers and they plan to do so until 2050s. PDP-11 is a 16 bit computer released in 1970. DEC, the company that produced them, ceased to exist in 1998.

At least the PDP-11 has error-correcting RAM and OS-level memory protection.

BTW, I wonder what they use for storage, these days... funny enough, the controller in any modern hard disk should be orders of magnitude faster than the PDP-11's CPU.

That's something I did a long time ago, but I wish for them, that their MS-DOS and Windows 3.1 run under a VM on a linux machine.

Except that DOS can be used in simple, real time systems, if on native hardware. That's because DOS has no kernel. The app is guaranteed 100% CPU time, aside from whatever interrupts might need to be serviced. So, we don't know if among the reasons it uses DOS is that it needs any realtime service guarantees.

 

[Geef]

My first tech job was secured when I mentioned I knew how to secure an ip address through Windows 3.11 and DOS 6.2 because I had the OS's memorized. Back in the year 2000 being an ISP's tech support who actually knew and used the OS's was kinda rare.

A lot of the techs didn't even know PCs that well, just enough to go through the steps to get a cable modem online.

 

[DavidC1]

The old software is so better written that it flies on a Pentium 4, 512 EDIT: KB MB ram, 32GB HDD compared to the newer stuff. It's response is near instant and the new stuff has a significant lag time and doesn't always respond. Apacs+ also is much more intuitive, keeps better logs and is easier to control. We've been migrating in pieces and the new software is so full of bugs.

Ugh, I agree. You read about how the Rollercoaster Tycoon game has been written entirely by a single person in 2 years in Assembly no less and runs on a Pentium 200 MMX while Blizzard devs are saying it's difficult to get inventory updated because of many players.

My impression isn't just because I have more time with that software, I have shown people trained on the newer systems how fast and easy the about 20 year old system is and they are in complete agreement, then I show them what it is running on. IDE cables and everything. The control systems are also isolated from the outside world, the software really isn't secure by any means. Probably the same with the rail software.

They worry about hacking on cars and computers for power grid. Why did they connect the power grid to the internet in the first place? Oftentimes, the IT people are the most stupid and cowards.

 

[tomscomments]

I live in Germany and could admin Windows 3.11 and DOS. But I'm happily retired.

I do have Windows 3.11 and Dos 6.22 installed in a virtual machine though...just to tinker with.

i spent some time in Germany some years ago, very nice country and nice people
you should help them

nb : i guess it is not for high speed train siemens trains however
Germans have many similarities with british. They like to keep old stuff as long as it works well. They love to keep old homes, old machines, old castles
you can see that while visiting small towns in both countries. Beautiful

 

[shady28]

At the chemical factory where I work we have been slowly been migrating from Siemens Apacs+ control software run on Pentium 4s because some critical hardware for it is no longer manufactured. We have some on Allen Bradley, but somebody decided to go with DeltaV with a modern server/thin clients setup.

The old software is so better written that it flies on a Pentium 4, 512 EDIT: KB MB ram, 32GB HDD compared to the newer stuff. It's response is near instant and the new stuff has a significant lag time and doesn't always respond. Apacs+ also is much more intuitive, keeps better logs and is easier to control. We've been migrating in pieces and the new software is so full of bugs.

My impression isn't just because I have more time with that software, I have shown people trained on the newer systems how fast and easy the about 20 year old system is and they are in complete agreement, then I show them what it is running on. IDE cables and everything. The control systems are also isolated from the outside world, the software really isn't secure by any means. Probably the same with the rail software.

The software to run the rail may be as good as it needs to be and getting the software/pc hardware/rail hardware to all match, all at once may be a heck of a job and might be why it has been put off so long.

Similar occupation. I have a bunch of sortation systems in multiple distribution centers, many still run off old DOS systems with STD Bus I/O from the 80s and 90s using custom written software in Watcom C. Those systems download tables over custom sockets and place them into memory in B-tree structures for high speed lookups on the business logic side. The 'PLC' is essentially custom, they can look up an entry in a 350K record in memory structure in under 20ms, similar to the scan time on a PLC. Anything over 25ms lookup and it shuts the equipment down, due to not knowing if it detected input states or not, just like a PLC fault caused by going outside its scan time.

We also have a variety of systems from Windows XP based controls software (old Steeplechase, and Beckhoff), windows NT boxes from the late 90s running some very custom HMIs to display running systems.

So, replacing one of the physical systems is anywhere from 3M to 40M. Upgrading the controls I/O and software software ranges from ~500K to ~2M. I have ~55 of these in total spread across 9 locations. We're building out a new system to replace a 40YO one right now that is ~35-45M.

The 'normal' IT world is clueless about these types of systems. I've had so many IT architect types ask 'when are you going to upgrade X', my standard answer is 'That'll be about 6-18 months and $1.5M minimum, get that approved and get me off project Y to do your upgrade and we got a deal.' They know they can't do that so they just spin in circles.

A control cabinet, just the hardware, can easily cost 250K. Some of these systems have multiple control cabinets. These things cost so much, take so long to put in, and are so disruptive to existing operations to upgrade, corporations put them in figuring on using them for 30+ years. This is totally normal. This is why you have firewalls.

 

[bit_user]

You read about how the Rollercoaster Tycoon game has been written entirely by a single person in 2 years in Assembly no less

That's just dumb. I mean, if someone enjoys writing asm that much, then I guess do what floats your boat. However, don't pretend for a second that makes a whit of sense. Even something like a video game will spend at least 90% of its time in less than 10% of the code, if not more like 99% and 1%. So, if you can't get good performance out of a C compiler (and believe me it's usually possible, if you really know what you're doing), then just optimize the hot spots or fast path in assembly.

I'm not even really big on C, because it lacks a good standard library of abstract data structures & algorithms. So, you'll often see C code getting bogged down because someone didn't want (or have the development time) to use a more appropriate algorithm or data structure. And while code optimizations can sometimes net you a small multiple of performance, algorithmic optimizations can deliver wins that are multiple orders of magnitude!

They worry about hacking on cars and computers for power grid. Why did they connect the power grid to the internet in the first place? Oftentimes, the IT people are the most stupid and cowards.

I'm pretty sure the US power grid isn't some monolithic thing, like you seem to suggest. From what little I've heard, it's a far more diverse and loose coalition of operators. Even if some are properly air-gapped and operated over isolated networks, you'll never get them all completely isolated.

 

[Jason_29]

MS-DOS 3.11 did not support TCP-IP nor DHCP out of the box did it? Wasn't there a Novell Netware client you could install?

 

[Rob1C]

Don't be sod

 

[shady28]

MS-DOS 3.11 did not support TCP-IP nor DHCP out of the box did it? Wasn't there a Novell Netware client you could install?

You had to use 3rd party TCP/IP add-ons. You can also use 3rd party add-ons to get up to 64MB of memory, but the executable had to fit in 640K. The extra memory was useful to hold data, like if you are loading an array of data structures with many records.

 

[carl_carlson]

You had to use 3rd party TCP/IP add-ons. You can also use 3rd party add-ons to get up to 64MB of memory, but the executable had to fit in 640K. The extra memory was useful to hold data, like if you are loading an array of data structures with many records.

This vividly brought to mind all of the old "download more memory" scams and just for a flash, I felt like I was right back in the "Wild West," i.e., the 90's.

 

[DavidC1]

That's just dumb. I mean, if someone enjoys writing asm that much, then I guess do what floats your boat. However, don't pretend for a second that makes a whit of sense.

No, it was in early 90's. The specs for RCT is 90MHz Pentium and 16MB of RAM. The overhead is reduced because it's more pedal to the metal, especially if the programmer knows what he is doing.

View: https://www.youtube.com/watch?v=ESGHKtrlMzs

 

[bit_user]

No, it was in early 90's. The specs for RCT is 90MHz Pentium and 16MB of RAM. The overhead is reduced because it's more pedal to the metal, especially if the programmer knows what he is doing.

I started writing assembly language on a 80386, so I know quite well what that hardware could do and how much you could gain by resorting to assembly language vs. C or Pascal.

I knew a few things about code optimization, too. I read virtually everything Michael Abrash published - if you don't know who he is, then it would seem you're the n00b, here.

The fact is, it never made sense to use assembly language, outside of optimizing the hot spots. Maybe if you're programming a microcontroller with only like 4 kB of ROM...

 

[aldaia]

I started writing assembly language on a 80386, so I know quite well what that hardware could do and how much you could gain by resorting to assembly language vs. C or Pascal.

I knew a few things about code optimization, too. I read virtually everything Michael Abrash published - if you don't know who he is, then it would seem you're the n00b, here.

The fact is, it never made sense to use assembly language, outside of optimizing the hot spots. Maybe if you're programming a microcontroller with only like 4 kB of ROM...

I started writing assembly for PDP-11 and later moved to i8086. At that time still made sense to write in assembly.

Unfortunaltely Michael Abrash was usually too late for me. His book "Zen of Assembly Language" would have been great in those early times. But the book appeared when we had already moved to VAX-11, 80386 and 80486 .

Back to topic. If having a few nuclear plants controlled by ancient PDP-11 looks scary. It's even more scary to know that all US land based ballistic nuclear missiles (ICBMs) are controlled by VAX-11 CPUs released in 1977

Rapid Execution and Combat Targeting (REACT): Armageddon with a Floppy disk and trackball! - Nuclear Companion: A nuclear guide to the cold war

The Rapid Execution and Combat Targeting (REACT) was an enhancement program of the U.S. Air Force. Its objective was to renew the SAC’s Minuteman missile launch control centers. The Strategic Air Command (SAC) initiated the program in the 1980s at the cost of 632 million dollars. As a result...

nuclearcompanion.com