What did Unix fans learn from the end of Unix workstations? When the market for proprietary UNIX workstations collapsed, few vendors survived… and those that did seemed not to learn much from it. OSnews has an interesting post on "the mass extinction of UNIX workstations," and makes a some good points. Some of them are interconnected: for instance, while a decade or two back, old UNIX …
I agree with so much of the sentiment in this article. Not just as a developer, but as someone who's fascinated with the history of the industry.
Would I love to run some of these in an emulator? Oh yes. Would I have a clue what to do with them? Probably not, but I'd love to find out.
Would I love to run Multics or ITS in an emulator? Definitely yes. Will I ever get the chance? Even more probably not.
It's a shame. This is history, being destroyed right in front of our eyes. I get the same sort of sick feeling I had when I saw the Taliban had blown up the Buddha statues as I did reading this, and I know that doesn't make much sense, but I hope you get the idea.
We need someone proselytising this, making things available, or at least pointing the way. *Sigh*
Icon in case some of you now need one, as I metaphorically do -->
Check out "PanelSim"! it's a front-panel simulator + SimH. There are software kits to run on it. I've got ITS v1633 running on mine. I've also got TOPS-10 v7.03. Getting TOPS-20 going is my next project. I've also got a bunch of PDP-8 stuff, IBM mainframe stuff (HerculesStudio with Michigan Terminal Service v6.0a, VM370 Release 6 + CP + CMS, and OS/VS2-MVS 3.8J.
In "smaller" systems, I've got a Data General Nova 4, HP-2116, HP-3000 an AT&T 3B2, (SimH), a TI-990/12, and a /4, and, going back up to bigger systems, a Univac I and Univac II, and Desktop Cyber (Control Data 6600-series).
It's all out there, waiting for you...
Ah thank you muchly. I tried SimH about ten years ago, I'd forgotten all about it! ITS and Multics I was vaguely aware people were scrabbling around trying to find all the bits, didn't know anything had come of the ITS effort yet, will definitely check that out too!
ITS is of interest to me too as I also didn't know about it. I've had success with TOPS-10 and -20 under SimH both of which were enjoyable and a bit of nostalgia: Hatfield Poly (as was) had 10 on its KL10 and 20 on its KS10 back in the day, imaginatively named BLUE and ORANGE. I regret never becoming fully acquainted (though an impossible-to-resolve admin cock-up meant I only had something like 10 blocks to play around with anyway) though the TOPS-10 machine still felt fairly ubiquitous as its FEP ran half the college's terminals.
Very interesting systems to explore properly under SimH, even though I really didn't have much of a clue what I was doing. I understand a version of TOPS-20/TWENEX/whevs was earmarked for the upcoming KC10 which eventually got canned, which I suspect to be the victim of in-fighting within DEC more than its supposed performance shortcomings: there was a lot of that about, especially towards the large systems. I still remember BAH (one of TOPS-10's former devs) ranting about "small computer thinking" on newsfroups years back and never quite understood what she meant, but I think now I have a better idea having read more about Cutler's attitude and general approach to stuff. Pity the KC10 never made it, it had some interesting features though it's hard to say how much of it was reality and how much was a wish-list; the idea that an option was native support for IBM-compatible bus+tag channels via IOPs was interesting (and seems quite likely considering how many 10 shops wanted it: it made sense for them to buy DASD using existing IBM contracts as they worked out a lot cheaper than the DEC equivalent, but bulky B+T-to-Massbus protocol converter boxes on every drive were annoying and probably gave worse performance) but with stuff like that and expanded SMP capabilities I'm wondering if half the problem is they were trying to be too ambitious and compete with IBM's Biggest Iron for a fraction of the cost.
Er anyway, I mentioned in reply to something else that I shouldn't start prattling on about 10s and now I have anyway! Hmm, a workstation based on a Kx10 microprocessor would've been interesting, though, and it's not as if the 10s weren't already acquainted with graphics.
Multics would also be interesting to see, as would CTSS, though I'm not especially optimistic about either happening; I'm not even sure there're emulators for the relevant hardware, though I admit I haven't looked. Same with ICT/ICL and George 3/4 & VME, for that matter, especially having recently discovered that OH's mother worked for ICT and wrote bits of the then pre-release 1900-series' libraries back in 1963/64... we were wondering how her parents met and figured it was probably through work (dad worked doing nuclear things at Harwell and a couple of his closest friends there were in the computer department which was involved with IC[TL] at the time) as they never mentioned it; probably nothing to do with official secrets, more a case that the past is in the past and even "but you were doing RL Tomorrow's World!" gets a bit blasé after a while.
Why am I still writing? I'll shut up.
Interesting you mention IBM peripherals on DEC 10 there. We had the Systems Concept SC10 on our dual-proc KI10 in Imperial's HENP (as then was) group. Only had tape on it, but it got us 6250bpi well before DEC managed to deliver, which was the main goal. That, and reliable tape drives.
Confused the hell out of the IBM FS guys when they asked to run OnLine Tests...
I'd heard that DEC's tape units of the era were less good than they could be, so that sounds like it was definitely a smart move. I hadn't heard of that particular one before but was aware that PDP-11s fitted with bus+tag were fairly commonplace parts of various IBM compute-o-spheres, such as MIT using them as FEPs for MTS's horde of ASCII terminals and other places finding they were handy gateways from their System/370 to other systems as well as TCP/IP and X.25 networks.
Didn't know there were multiprocessor KI10s out there. Wish I'd seen all the blinkenlights on that! I think the KI could have some sort of record for the largest festoonment of Christmas-tree lights once you include the memory cabs; it's a close run thing with the bigger 360s like the 75, 91 and 195 but I think the KI10 might possibly win that round.
Oh yes, the blinkenlights were a joy to behold. Getting the DEC engineer to replace all the faulty ones was always a challenge.
But.... when we had to take the slave off maintenance 'coz of budget cuts (late 70's UK IMF rescue days....) I got to recognise the logic state when we got SSP (Stop Second Processor) crashes. Apparently random series stabilisers (per row of TTL logic) tripping out on overcurrent. Finally (after 4 pints of Directors one lunchtime) tracked it down to an O/C end winding on the PSU transformer, so the unregulated DC feed was a volt or so down. Dropped it down a tap each side of the halfwave, all was good, and DEC none the wiser when it went back on contract.
We used TOPS-10's brilliant ability to do both timesharing and realtime for the data capture from an HPD flying spot digitiser for bubble chamber film. Although timesharing did stop for a couple of seconds while a frame scan took place.
I remember the joy when our systems programmers finally got SMP going, that was the time the second processor started earning its keep.
>I'm pretty sure most corporations believe 1 recovered machine = 1 lost sale or 1 recovered machine = 1 lost cloud subscription.
I think MS would regard anyone running anything less than Win11 with M365 as:
n lost sales + m lost cloud subscription
Where n = number of subsequent Windows releases and m = number of people sharing that machine.
It's possible to emulate IRIX on MAME, but don't expect snappy performance. It's pretty painful in fact, but is certainly an impressive acheivement. Would be great to see more of these emulated in due course. It's always difficult though when both the underlying hardware and software are so proprietary.
This post has been deleted by its author
Most of that is down to scale. Modern systems are physically very tiny, with almost no thermal mass. So they quickly get too hot and stop working without cooling.
Many of these older CPUs would run at their maximum pelt without any heatsink at all for quite a long time in 20C ambient air.
You also notice a ramping fan much more than one that runs continuously, and fan speed control is relatively new in desktops.
I do have, squirrelled away, an old NCD X-terminal. Square form factor, 1024x1024. Only 15" I think, and mono, but so crisp and clear compared to VGA PC monitors of the era. Sadly only runs X11R4 in ROM, no idea whether there was ever any way to upgrade it or network boot it to a newer version.
$FORMERJOB used to have about 100 or more NCD X-terms as the engineering staff were using them for Framemaker documents and software work. They even had colour models.
(It was cheaper to have NCD X-terms and a few big servers than pizza-box workstations for everyone.)
If the electricity was interrupted though, network booting 100 X-terms took a loooong time.
I gave a presentation at the local DECUS conference about how we used them, how to setup network booting etc etc. One question I got: "Isn't a PC faster than an NCD? Why not run Windows plus an X11 software package?"
I have to say I was surprised. It hadn't occurred to me to compare the two. My answer? "Not this year, but maybe when PCs hit 450MHz they will be competitive."
NCDs only lasted about 3 or 4 years in the marketplace, as I recall. PCs + Hummingbird X-Windows software replaced them.
"PCs + Hummingbird X-Windows software replaced them."
I had pretty good luck with DesqView/X and the optional Motif and TCP/IP packages. Worked quite nicely with the existing (mostly) mixed DEC and Sun networks I was playing with at the time, but was a trifle spendy.
But then, coming from a BSD background all similar commercial products seemed a trifle spendy ...
We had lots of NCDs at a former employer, mostly 19" B&W models. They were incredibly sharp but I found their 1-bit graphics quite hard work and managed to wangle myself a 17" colour model. Lower res (I don't quite remember after all these years, but I think the 19" screens were 1280×1024 and the 17" was 1024×768) but it was colour! The last model I saw was a then new 19" (or maybe even 21") colour with an 88000-powered pizza box under it instead of the more familiar 68K plinth. Fancy, but by that time I'd "acquired" a VaxStation 3176 with a 19" Trinitron monitor and SPX graphics (1280px I think)... both of which are still in the garage. The Vax is poorly (bad serial port according to the diagnostics; probably just a dry joint) and I'm not sure I'll ever be able to move the monitor again, I could barely lift it when I was half my age back when I was a lot fitter; and my internally-generated VMS licences expired 20 years ago, which is a problem as nobody seems to be able to agree who owns VMS for Vax nowadays. :| I suppose I could run NetBSD on it if I get it fixed, though it'd only be for the novelty value as it consumed a lot of electricity in its heyday (as I discovered when I participated in DEC's wfh programme in the '90s) so I figure I'm better off sticking with SimH anyway. Er, I think I've gone far enough off-topic without starting to blather on about PDP-10s as I was about to do...
I was given NCD xterminals to use in a former job and I did like the large screens and superb resolution but they were 'dumb clients' and I preferred to have some local smartness and a computer onto which I could install my own choice of operating systems and applications and keep working when the central server crashed (which it did quite often).
YEARS ago, I started work at a small company that used ViewLOGIC schematic capture software on 486 machines (I told you it was years ago -- 1994 at a rough guess). This seemed like a great idea, until I discovered that the system would crash if I had more than one page open. This was a QEMM problem, I believe, and I could see that it wasn't worth my time to mess around with it when there was another alternative.
Rather than fix that problem, I brought in a second HDD with early Linux on it, installed it in the PC and dual-booted to it, opened an Xterm and remoted it to our Unix system, which also had ViewLOGIC on it...the UNIX version, which didn't crash. The rest of my time there, I did my schematics on the PC, configured as an X display, with the actual application running on the Unix system.
It was about this time when I realised that early Linux was a far better OS than Windows 3.1...
Hummingbird X-Windows software
Oh God, how bad that was!
It was so infuriatingly bad that it drove several people, self included, this was "way back" when there were barely any distributions available. I think there was Slackware and that it was a swine to get working on a 486.
I had 8 or 9 ncd xterminals to support in the late 90s-early 00s. Needed BOOTP and TFTP servers to boot, configure networking, font servers and grab any local client software (via nfs?) The server(s) were DEC Alphas DECOSF/1 3.2G then DECUNIX 4.0B I think.
The users could choose from a list of shared DEC, HPUX, SGI IRIX and IBM AIX machines. Given each had a different desktop environment I would say users were made of sterner stuff back then.
By the time the Unix boxes were supplanted by Linux x86_64 boxes the terminals had disappeared. x86 32 bit Linux (or Unix) was pretty useless in that scientific computing environment.
I still have the NCD cdrom that came with the last terminals squirelled away with irix, hpux, dec osf etc and sunos/solaris cdroms - the rs6000's aix software was on cartridge tape so never bothered.
Great days? Perhaps not - supporting six different architectures, 32 & 64 bits, big & little endian, diferent major versions on what, by today's standard, was quite peculiar hardware - challenging is probably an understatement.
Ironically with the plethora if Unix platforms, software was generally written far more portably than seems the case today. (Of course I mean open source.)
I never thought I'd see it but Adobe PostScript source has been released into the wild.
@Arthur the cat Wow that is unexpected. Postscript is not obsolete printers are still being made that support PS.
PS was always a requirement for new printers at my until recently employer, as our software supported PS along with many other protocols. In fact we coded our own drivers. I remember writing the C code to generate PS page for page rotations along with the other protocols.
I remember coding PS to generate some figures (mostly because I was curious). I also did manually edit some figures so the print server would not let all points drop thruogh the raster in teh rastering process (we ended up missing a figure on a conference poster, about 20 years ago).
I remember having the Adobe "blue book" -"PostScript Language, Tutorial and Cookbook"
There was also "PostScript by Example" by Henry McGilton and Mary Campione
My first exposure to PostScript was a friend getting a "zine" from an FTP site written in this newfangled "display language" that was supposed to be the same no matter the printer. I don't even remember the subject matter, but I do remember the struggle to get it printed and the awe at the clean, fancy, and pretty final product.
Why the hell doesn't El Reg support the <u> tags for underlining titles?
Some companies are remarkably stubborn about this. I wish IBM would do a hobbyist licence for VM/ESA which Hercules folk have wanted since its inception but they won't budge. Makes no sense, you'd think they'd welcome the likelihood of it having a (much) bigger profile and it's not as if anyone's going to try undercutting their mainframe business by running an enterprise-level system on a 40-year-old OS under PC emulation...
Considering they were pretty quick to open source their office suite, it's doubly frustrating. Cost of unpicking the legal mess presumably means it's not worth doing it unless some IBM Lawyer decides to do it as a pet project.
The last release or two of SmartSuite was available for non-commercial use for no cost. I remember several magazines carrying it on their cover CDs back in the day (and I still have some of the CDs somewhere).
I much preferred WordPro, Freelance and Lotus123 to the then current Microsoft products.
But it was never released under an open license, merely a permissive one.
IBM no longer own this software. All of the Lotus derived products made their way to HCL a year or two back.
In many ways VM/ESA is still leading-edge. The 390 architecture might be increasingly niche, but some of the things VM could/can do in terms of nested virtualization, are still things that more "modern" architectures struggle with.
Back in the day, licensees of VM got most of the OS as rebuildable source - modifying the HCPBOX module to create one's own EBCDIC-art login screen was the first thing one did when installing it on a new system - but some bits of it were object code only (OCO). I always assumed that was where magic happened.
I remember Lynn Wheeler describing some of the stuff they did in general terms, right down to counting the path lengths in cycles to work out what they could shave off here and there. Still active (on FB last I saw) and writing interesting stuff. Ironic how IBM always regarded VM as the unwanted stepchild and to a significant degree it still does compared to golden boy MVS or Z/OS or whevs in spite of it being such an important part of its mainframe line's ongoing success.
I once had a book. quite apart from the official collection - Red Book, Blue Book, Type 1 format book etc, that was pretty much an annotated disassembly of all the key bits of the first PostScript interpreter. Would have been early 1990's. A vital book if you were doing any very low level PS programming as there were lots of interesting bugs in the first few releases. But as someone who learned Forth in the 1970's while hanging out in an observatory I never found PS programming much of a problem. Now working around the bugs, that was another matter..
PostScript is a rather convenient way of creating documents when coding applications. Run it through PS2PDF and you have something you can email.
If you want to create PDFs directly an understanding of PostScript makes like a little easier.
I own a number of SGIs, lovely machines. In fact, I learned to write C++ a couple years ago by writing a network program that lets you control an IRIX machine that is connected to a capture card, using the mouse and keyboard of a Windows machine. https://www.youtube.com/watch?v=lOhT0x5m9Nc
Now I am a professional C++ dev. So yes, learning to use this UNIX workstation directly impacted my career.
Same here. Have an old O2 I acquired for free from a local university when they were being retired, subsequently picked up an Indigo2 and a Fuel. Beautiful looking machines for their time and I thoroughly enjoy firing up IRIX from time to time for a bit of nostalgic fun, but finding spare parts for them is becoming increasingly difficult and expensive sadly.
I did used to have a couple of Indys, but sold them ages ago as they weren't being used, plus mine were fairly modest spec and hence quite slow. The O2 was essentially the successor to the Indy. Not the snappiest system, but they do have the fun of having a unified memory architecture (no separate system and graphics RAM), which means it's possible to use video files as textures in models just by passing the relevant pointer. Quite radical for the mid 90s!
I will agree however that the plastic cases of them are unbelievably fragile and brittle.
There were more than a few O2 in the labs for dev work which didn't have the main upper skins at all -- most of them had (at most) the lower round-ish skirt, and the boxy square metal chassis was simply bare.
You also didn't see the same kind of homegrown creative variety in O2 paint schemes as happened with Indigo and Indy. Even putting aside some of the prototype primer/paint jobs that showed up in the labs for Ind*, people would sometimes customize the chassis on their own. I've seen a whole rainbow of Indy chassis lids, presumably those were favorites because the 1-piece was easier to paint than Indigo, I2, and O2 multi-piece cases.
I liked O2 just fine, but I miss the Indy more. We used those (and Challenge S) for all kinds of stuff....
My primary workstation in the 90's was an Indigo2. I really loved the interface. Showed how Unix (IRIX) could look amazing and so much nicer than CDE in my humble opinion. The end came quickly for SGI as a large-scale computer vendor. The disastrous move to Windows was the final nail.
I have an Indigo. Got it from my old boss, who was using it as a footrest after it was retired from being a CAD workstation. Got bored one day and decided to see if I could resurrect it. About $250 later (VGA adapter, AUI to 10BASE-T adapter, AT keyboard and mouse adapter), and after cracking the root password, I got it working.
Updating the OS required using my Linux system as a TFTP server, since the system I got doesn't have a CDROM. But the OS is up to the last version that would run on my machine, and all I'm missing now, is a C compiler (IRIX seems to be the only UNIX which doesn't come with "cc"). Understand that gcc can be shoehorned in, but I haven't had the motivation yet...hoping there's a version of the official compiler that will become available.
It works, it's lovely old Unix, and the demo graphics are fun to watch.
Footnote: The Computer History Museum is located in SGI's old HQ building.
I don't agree. I do feel nostalgia for my old computers, but they were ancient crap and modern crap is much better.
Workstations were just high-end personal computers, before IBM co-opted "PC". I learned UNIX on an old Sun pizza box. That machine was a beast at the time, but it's tiny compared to my current machine. SGI did awesome graphics, but my middle-level nvidia card blows it away.
I don't pine for old tools. I might want to keep around an old hammer, but that's because my Dad owned it. I do watch Vintage Machinery on YouTube, but that's because the old kit *is* demonstrably better than current stuff, and Keith Rucker is rescuing and refurbishing it to modern standards.
Of course old UNIX kit is worth a ton! So are old toy cars and Star Trek model kits! That's how nostalgia works. I had a roommate in college that kept a PDP-11/34 running, complete with a couple of RL-02 drives. He spent more time chasing up broken point-to-point wiring in the backplane than he did being logged in.
Linux has the same Windows-ish desktops all over, because that's what the developers know, and they don't know anything else, so that's what they write. CDE had the "advantage" there wasn't anything to crib from.
Fucking Sun NeWS. That was so very nice, but Sun decided to smother it under a blanket. It's one of the few computer things I feel a genuine loss for. One of the many reasons I'm glad Sun died horribly.
I still have to use Solaris at work. It makes me so happy I have Devuan at home. With Solaris, I feel like I'm using sandpaper wrapped around a block instead of a battery operated high power orbital sander. They do the "same thing" but one is so much better to use and does such a better job.
And YES, at least once a month I regret giving away my TRS-80 Model I (my first computer) and huge range of Atari kit. But that's just nostalgia. So is keeping track of the fact I'll have gotten that TRS-80 43 years ago this Christmas.
The main reason for me that (significantly-)older computers are so much fun is not running the OS (much of which was primitive, or even crap) and apps, but writing programs in assembly language an running them.
This is because the "older, primitive" architechtures were simple enough to be easily comprehended. First Zilog, then Intel started adding instructions like crazy, and all the other surviving manufacturers did the same.
About 100 pages of instruction manual/architecture manual are the "convivial limit". Modern x86 doc sets are multiple thousands of pages. And my brain isn't big enough to hold all of that ... nor even just the indices of all that.
I had an OSI Superboard (what was the UK board that was a direct rip-off of it?) and got to the stage when I could remember enough of the most common 6502 opcodes to patch code without needing the assembler. In fact LDA being 0xA9 is still floating round my brain.
I was recently reading about the IBM 360/195, an early and somewhat credible attempt of theirs at making a supercomputer, probably at the behest of former Stretch users/designers; and actually didn't do too badly at it either, in spite of the lack of virtual memory, it outperformed the newer 370 line until the '80s IIRC (and I think even the 308X/3090 only did so by merit of their much higher clock speed and more memory rather than cleverness). Anyway, it was complicated by the fact that unlike other 360s (and 370s for that matter) it did stuff like out-of-order processing and what-not, but even so, the ways of dealing with it, both taking best advantage and avoiding some of the gotchas which potentially broke the 360's POO (the imprecise interrupt being a particular headache) were still reasonably straightforward and concise, so although optimisers were the usual approach, assembly programming was still feasible and was very much a thing. Rare that you see it these days though, certainly outside of stuff that can't be done in C.
There may be thousands of pages of doc's in the current Intel x64 manuals but you can ignore almost all of them. Only about twenty or thirty are relevant.
Now the Intel instruction set has been a train wreck since the 8080A days (let alone 8086) so my old rule still applies. For the first few decades it was ignore segments and any x32 instruction that did not easily map onto 6502/6800/68k. Because they are just bloody stupid. And for the last few decades its ignore any instruction that does not map easily onto 68k/ARM. Saves a huge amount of time because most of the x32/x64 instructions are irreverent, added by marketing t o make it look like they have actually done something new. Unless you are writing a SSE/SIMD library there is never any need to venture into that swamp. Just use a library. Let someone else go though that mental anguish instead of you.
I can think of only one improvement in the Intel instruction set in the last 40 years. You can now do pc relative code. Something which every other non Intel processor could do 40 years ago. God, I hate Intel. If only the trace of the San Andreas went through Santa Clara rather than up in the hills...
1. How do I locate the parts I do NOT want to ignore without reading the whole damn mess? Oh, and don't forget the "errata sheets" (descriptions of things that don't work the way the other parts of the manuals say they work).
2. Can you bring a modern x86 from power-up to a useful state if you ignore all the write-only, Machine-Specific Registers which set up ... "stuff"? Does the BIOS do all that dirty work for you?
I quit following Intel-architecture assembly-language details when the 80286 came out.
1) Thats what the TOC are for. If you know how to program asm, any asm, working out what you need to read is straight forward to do. Mostly its a New Stuff, Ignore..scan read.
If not an asm programmer then learn a proper instruction set first (ARM, or even MIPS) then a quick look at any programming x86/x64 book should give you fair warning if what you are in for. Then look at the TOC.
2) Your mileage may vary depending on the BIOS and chipset, and just which x86/x64 CPU you are dealing with. If its a SOC you read the vendors docs first. Anything more complicated the chipset vendors docs are the place to start with. If it comes with a motherboard, then the vendors docs first. Then you look at the Intel docs TOC.
I've pretty much totally avoided the Intel universe over the decades in my asm programming career. Just learned to read the disassembly mostly when digging through crashes. Last heavy duty x86/x64 was writing a codegen output for a compiler. Normally you just fill a table with basic opcode bit pattern and bit twiddle bytestream output as needed. Not with Intel. Oh, no. Nothing so easy. Totally bloody stupid. At every level. All the way down.
Now with the ARM codegen, everything is where you expected in the binary map of the instructions. Very straight forward. Every decision in the binary layout of the instruction set was reasonable in how they did things. I like ARM. And every other instruction set not architected for marketing by technical incompetents. Remember, the 8080 / x86 architecture had already been obsolete for almost one / two decades the day they shipped their first chips. A technical dead end abandoned by mainframes and minis in the early 1960's.
They dont all suck. Some are actually very pleasant to use. Some are OK. And some are just horrible messes. But there is only one Intel x86 Lowest Circle of ASM Hell.
The 68K was a pure joy to write for. MIPS and ARM both very pleasant. As was the Transputer instruction set. The 6502, 6800, PDP8, PDPII, / VAX, PPC , were not bad. Z80, SPARC and SH8 could be a real pain at times. But the x86/x64 is just a total train wreck. Pure torture to use. Stupidity piled on stupidity.
And I could add about another half dozen or more instruction sets of more obscure CPU to the list. All in the OK category. I only once saw an instruction set even more bizarre than x86. And the was for some custom Japanese micro-controller back in the early 1990's. Where it ended up as the defacto half assed video controller in a game console.
"SGI did awesome graphics, but my middle-level nvidia card blows it away."
Well, obviously. The 3D performance of my current mobile phone blows SGI graphics capabilities away, but you can't really compare modern graphics to those from 25 years ago. For their time, they were fantastic. By today's standards, they're positively prehistoric. But then, people don't use old Unix systems in order to be impressed by their performance compared with modern kit.
Back in the late 90s I had a curious one-off contract at an airline pilot training centre. They trained the pilots in a Boeing 707 cockpit mounted on hydraulics. The whole shebang, including flight simulation, was controlled by a bank of four PDP-11s. It had been working seamlessly since the late 70s, but the PDPs were getting a bit long in the tooth and broke down frequently. The software wasn't broke and didn't need fixing, so they just wanted to port it to more modern kit; unfortunately however, the original (FORTRAN) source had been lost in the mists of time. The solution was simply to write a naive PDP-11 emulator, which we did, in C. It ran flawlessly on a single off-the-shelf 80386 PC with about 64MB of RAM - in fact we had to clock the emulator pretty slow to sync with the hardware.
Back in the early 90s, a guy at my local Amiga user group had a Canadian CF-100 instrument trainer from 50s in a trailer in his yard. It was essentially a cockpit that you climbed in and they through a blanket over to simulate bad weather night flying. Just listeni to "tower" for headings and watch the beacon heading. He had a picture of the original room-sized computer that ran it, which he replaced with a C64.
Thanks for reminding me of that!
I agree 110%. Nostalgia is wonderful to visit, times and the devices that occupied space all seemed a lot easier back then. That always tickles our brains, "simpler times", with devices we can seemingly all get a mental grasp on.
But after you waste time fiddling with devices of limited capabilities in today's world...you need to get back to today's world, and actually get things done.
And those nostalgia devices will almost always let you down here.
The past is always a nice place to visit from today, what would be their future. It's charming and easy, because we've only ever gone more complex in life, never easier. But to try to live there is dooming yourself to a life of both eternal stagnation and frustration, as you attempt to impose the complexities of today on creations that weren't designed for it. I have two friends stuck in that past, one who listens to his multiple Victrolas and 78's, but then laments how come his dreams of finding the perfect companion have never come to fruition (as he searches based on those dreams, rather than the realities of life around him).
Nostalgia can be a poison as you idealize a past...that never really was.
I think we can see the "good" elements of the past without ignoring the flaws and shortcomings of the past: racial/sexual/religious discrimination, heavily-cast/enforced role models, poor-compared-to-modern-times medical care, etc.
> I had a roommate in college that kept a PDP-11/34 running, complete with a couple of RL-02 drives. He spent more time chasing up broken point-to-point wiring in the backplane than he did being logged in.
I worked with PDPs for a number of years and never once has backplane wiring problems. (Even made a change to the backplane wiring to support a static RAM card for an 11/70---no problems.) I have to wonder what the heck he might have been doing to create those.
A painful part of my career was filling all available space with obsolete ex-spiffy hardware, and then running it out to disposal when I couldn't turn around any more. It was astonishing how much silicon and tin and CRT a small low-tech academic department could generate. Yes, I stripped it ruthlessly. RAM from three NeXT machines went into laser printers. 386 machine case and power supply could fit a 486 motherboard. Even so, every year I filled the 1979 Thunderbird; or joined with a neighbor department and borrowed a box-truck. I've kept a few tidbits for personal amusement, but I can't regret scrapping even the exotic MO drive in the NeXTs, because The Organization provided NO support for even minimal storage.
These days I keep a 2009 NetBook for a card game; everything else is recent.
I dabbled into a Dell R210* for my TVheadend server (yrs ago) obviously a WEEE resale. Prob was the techs at orig owner removed HDs by removing ENTIRE HDD cage/mount
Absolute bu**er to find and sh*t costly to buy (even on eBay)
*Less power hungry and easier on the eye in my basement than orev HP DC7800 CMT
I made an AT case out of plexiglass (perspex, acrylic, whatever) once, in an attempt to make computers somewhat less inscrutable. Nowhere did I.need copper tape, everything was properly grounded with the ground lines in the cables. Yes, it was a trifle noisy from an RF perspective, but I rarely fired it up ... it was more a visual aid when teaching newbies.
About 15 years ago, I picked a full NEXTstation from my former university, a friend there told me they were getting rid of the kit.
Got it running (once I got the special pizzabox to display cable I had forgotten), installed NetBSD, then tbe last version of NextStep, or was it OpenStep? Installed Doom, of course, to at least run the original version once (black and white). All nice fun!
Then a decade ago, I needed the space, and bartered it for a bottle of wine.
I think the worst thing is having to get rid of your "collection" for domestic reasons. I guess everybody's been in the position of having to move, or downsize or something else and the old tech is just taking up too much space. Then a few years later you come across the missing piece, or someone puts the software or the service manual on line... sigh.
I know the feeling.
I've rescued and then donated or otherwise given away so many 90's era Sun, SGI, and DEC kit over the years, I can't keep track. I kept and enjoyed some of them for a time, but eventually all were passed along to someone else. I like to think a lot of it is still out there running.
And really, that's the most you can do: try to find a good home for the gear if you possibly can. and hope that the next person down the line will do the same.
Some resources I used successfully to donate old Unix systems:
- The Rescue list at Sunhelp: http://www.sunhelp.org/
Lots of folks willing to share and help, not 100% Sun but that's the main focus. I found good homes for lots of my old kit this way.
- The NetBSD ports mailing lists: http://www.netbsd.org/mailinglists/
E.g. "port-sparc@netbsd.org" and others. Because NetBSD still has at least some level of support for lots of old kit, this is a good place
to contact devoted folks for that platform ("port" in NetBSD terminology)
- OpenBSD, FreeBSD, and Debian mailing lists
Somewhat similar to NetBSD above, but with not quite as much old hardware support. I gave some DEC and SPARC64 stuff to FreeBSD
folks, but support for those platforms has diminished over time.
Probably a good idea to lurk and or browse the mailing list archives a bit before jumping right in. That's good advice in any case.
I was given a complete Lisa system -- including software and the box it all came in -- back in the early 90s. by my then Engineering VP, I guess as a sort of booby prize for recognizing what this oversized desk ornament was. I ran it for a bit with a conversion BIOS that made it into a Maclet but eventually someone -- "she who cannot be named directly" -- made me get rid of it. I wouldn't have had it today, though - its the sort of junk that ends up in a collection and now is probably worth big money.
I can't get nostalgic about old workstations. They were good for the year but Heaven Help You if you needed to change a disk. The Sun product was not only beyond expensive but the whole rigmarole of rebuilding the software was definitely not user friendly. PCs rapidly overlook them in performance and so became the better option.
So........I bought my Osborne 01 in September 1982....
...and most, maybe all, of the files written on the Osborne are saved today on my Fedora 37 server.
...this means that my Wordstar letter to my boss, resigning from my job in December 1982, is still available here at Linux Mansions. You know.....DOSBOX, an archived version of the MSDOS version of Wordstar....and.....voila.....reading my resignation letter from forty years ago.
And I've still got copies of the Osborne CP/M-80 software. I suppose I could run all of that (Digital Research tools, Wordstar, dBase-II, Supercalc......) under a CP/M emulator....but I don't have the time. Still....I guess with an emulator and the archived files, I could relive my Osborne days using F37.
Nostalgia......passes the time, doesn't it????
Heck, speaking of ancient data... my banking details on my phone range back to 1999, when I was strongly encouraged to buy a Palm III by my boss, who thought it would help get my life somewhat better arranged.
I've since ported the data from that initial banking app through the Palm range to the Tungsten T5 and TX (I went through Palms like candy) then to a Nokia N810, on to my Motorola Droid, and to my current Pixel 5a (and I've gone through Android devices like candy, too)
Man... that N810... Nokia had NO CLUE what the hell they were doing. It was not an upgrade from the Palm, but by that time, Palm had fallen on their sword.
>...and water cooling (to keep it quiet)
Obviously Intel - my equivalent year old AMD Ryzen 5 system (*) running 4 x 4k screens (it's used for video production) is whisper quiet without the need for water cooling or forced room cooling/air conditioning...
I think this is the reason why we saw the demise of the proprietary Unix workstation; they couldn't keep pace with the price/performance of the high-end PC, which at the time also ran popular office tools like MS Office et al...
(*) It replaced a dual Xeon Windows XP Pro 64 workstation with fans; everyone in the house knew when that system was running...
Silicon Graphics workstations. I regret not picking more of them up in the early 2000's before prices went insane.
IRIX I never had any special affinity (just-another-AT&T SystemV for the most part) for but the build quality, aesthetics and power of those systems for the time they were built are another league. A top end Threadripper Workstation today isn't the "leap" in power over it's desktop brethren that SGI was.
> A top end Threadripper Workstation today isn't the "leap" in power over it's desktop brethren that SGI was.
Depends on what you are doing, ie. whether the applications you need to run can usefully use all the cores, and what other components you marry it with. Eg. for video editing, a low-end threadripper system (or just a plain Ryzen 5 or 7) with a couple of high-end video cards/GPUs can outperform a high-end threadripper without video cards/GPUs.
But in general I would agree, however, for many there is a world of difference between spending sub £1k on a PC and £3k on a workstation.
"There are multiple forms of value in keeping old machines running: educational value, in seeing how things were done. Lessons in performance optimization, resource usage, scalability, and so on."
No truer word spoken in a long time on this topic. The sloppy engineering-free approach to modern systems that relies on massive hardware resources and performance to conceal appallingly inadequately written code, despite making life easy for 'developers', will one day (and possibly quite soon) reach its inflexion point leading to effective collapse of technologies we have come to rely on for societal survival. The signs are already there for those with the unbiased vision to perceive them.
I didn't know my kitted up U45 from long ago was worth that much.
It's in running condition, and I have the relevant Solaris Installation disks.
( no I won't sell )
I know where to find an UltraSparc 2 ( that has not been powered up for more than 15 years now ) with Multipacks and Unipacks containing various stuff ( from CD drive to Tapes and HDD ).
Sadly I got rid of the Sparc Classic when the NVRAM went AWOL and you had to go Forth in the bootrom to tell it where Solaris was hiding.
[Author here]
> Sadly I got rid of the Sparc Classic when the NVRAM went AWOL and you had to go Forth in the bootrom to tell it where Solaris was hiding.
Yeah, I gave away my SPARCstation IPX when it did that, too.
It is easy enough to fix, though.
Obsolyte documents it well:
http://www.obsolyte.com/sunFAQ/faq_nvram.html
And someone makes a replacement module:
https://www.glitchwrks.com/2017/08/01/gw-48t02-1
I'm still working on Sun Blade 150s and Ultra 10 & 60s in a production environment. These machines went into service in the late 90s and nearly 25 years later they are still doing their job. Very reliable bit of kit. Rarely have any problems other than replacing the NVRAM battery after a decade or two. The Blade 150s have the edge on reliability though Vs the Ultra10 with their ATX PSU which occasionally do fail.
My first proper job involved looking after loads of HP 9000 712 workstations running HP-UX 9.x and then 10.x with the HP VUE front end (which was some re-skinned bit of CDE or something). I learned quite a bit working on all that.
Then in my second job I ended up looking after loads of Sun kit from a fair few Ultra 10 workstations up to the stupidly obscene Ultra Enterprise 10000 monstrosities. At one point I also had an Ultra 10 and also briefly a Netra V100 at home, but never really used either of them for anything and must have given them away to someone who did have a proper use for them.
But with the demise of Sun and rise of Linux nowadays I can pretty much do everything from an SSH session and a web browser from a multitude of different devices.
We developed the cabin software for the Airbus A330/A340 on HP9000 workstations as both they and the aircraft systems were 680x0 based. I quite liked the VUE front end and made a version of the workspaces (virtual desktop) feature for RISC OS in the early 90s.
A few years before that in the 3rd year at UCL we finally got off the old Prime mainframe and gained access to a lab full of SUN 3/50s with large mono monitors and optical mice, which was great until they let the 1st year students log on from the green screen terminals upstairs and run Scheme or Miranda, which caused them to swap over the (thickernet) network, effectively freezing X for minutes at a time.
My last job at an academic facility had an e-waste bin by the loading dock of the department I worked in. An old professor there retired, and they got rid of a massive amount of old hardware he had been holding onto at the office. I snagged a couple of old Sun Ultra 60s but literally did not have room to take anything else with me. A lot of interesting hardware passed through there as the dept had been doing computational work for decades. Luckily the they had a policy that anything in the bins was free game and I think some was scavenged by other enthusiasts.
I had an Ultra 60 that I wound up getting rid of. Interesting to keep up-to-date on Solaris as we were transitioning to Sol10 at work but, to prepare for an upcoming move, it went out the door. It was a real power hog, too. I swear I could tell by looking at the spinning disk in the power meter on the back of the house that the Ultra 60 was powered up.
Blimey, looks my collection of Alphaservers/Alphastations will be worth something one day. Also my full sets of vms and tru64 software.
I realised there was $$$ in this when my VAX mouse sold for £50 in 5 minutes.
The physical hardware beats the emulator every time. However, getting the software is trickier these days.
I had saved an "Open" VMS7.2 Alpha distribution kit (the bookshelf box type). Looked it up on E-bay and wasn't worth the postage to send it anywhere. So I tossed it out.
This was in June this year.
But I did used to have an Alphastation, I think it was 133MHz and 64MB. We had used it as an NFS server for several cabinets of 2GB disks -- it was way faster than the competition (it had 100 Mbit FDDI) looped to a few SGI and Sun servers.
Quote from OA
Interesting article. Just a nitpick. Quote from OA...
"If half a dozen bitterly opposed vendors could cooperate to create CDE, how come now each major Linux distro has a dozen very similar desktops?"
Linux distributions package a range of software from projects upstream. The duplication of effort and the 'herding' to MS Windows style or MacOS style desktop presentations is to do with the upstream projects not necessarily the distributions. Upstream projects are just, you know, open source projects. Some are huge and some are just one person and a github account. There is no management committee to wield the big stick and demand rationalisation.
Of course, RedHat sponsor Gnome and Gnome is default in RHEL and clones. And there are smaller distributions that make a feature out of a desktop (Mint, the Chinese ones like Deepin &c).
A recollection: Back in the mid 90s I was teaching short courses in evening classes on how to make a Web page (html/Dida and similar and how to upload to web space using ftp and such). The students were from a range of backgrounds. I found that asking if anyone had ever used UNIX or had heard of WordStar made things go much quicker... UNIX systems were widely used in 1980s especially in local authorities and other public sector employers in the UK.
Icon: BASIC on the teletype into the local mainframe when I was at school
The "problem" was that vendors of hardware and software lost control of the PC. First, IBM lost control of the PC when the BIOS was cloned and then Microsoft lost their monopoly on the OS for the PC. They're still trying to get their control back with "Secure Boot" but IMHO that ship has sailed.
Suddenly there was real "competition" in the PC marketplace. Intel had to develop and push to maintain market share, PC vendors needed faster CPUs to attract buyers, and the impetus from BSD and later Linux led to "commodity microcomputer" hardware that was rapidly exceeding the capabilities of the proprietary UNIXes and the minicomputer and workstation hardware, and at a LOT lower price point. So the practices of the vendors of the minis and workstations started to look a lot like "rent seeking" and they very quickly saw their lunch being eaten by vendors with whom they could not possibly compete.
Why spend tens of thousands on a "workstation" when a clone PC has the same or more power? Why spend thousands on a gen-you-wine UNIX when Linux is just as good for about 99% of use cases? Sure there are still some specialized apps for niches where specialized hardware and software is needed. But with the GPUs and other high-performance PC hardware available and the ingenuity and hard work of FOSS developers, a lot of the vendors of the proprietary niche stuff available are going to be out of luck sooner rather than later.
Twenty years ago, I saw a "Nortel phone switch" that had room for 6 lines. Nortel was a big Canadian success story, and they pretty well owned the market for high-end telecoms in several countries. That phone switch had some hardware to handle phone lines and a very low-end 486 processor. The thing was over 10 thousand dollars in 2001 money! Sure enough, someone in China decided to built cards for PCs that could handle phone lines, and some bright sparks wrote Asterisk. Presto! No more need to deal with Nortel. And where is Nortel now? Long gone.
Some company in the middle east tried to sell us a WAN simulator for about $20,000 to simulate a WAN with controlled packet loss, delay and latency. I built one using FreeBSD and DUMMYNET on a scrap PC with a second NIC. Cost? 0$. Benefit? Saved 20 grand.
Screw the rent seekers.
I agree to a point. I still contend that trying to deliver an expensive and proprietary product while competing against FOSS and commodity hardware was and remains a doomed business model.
Perhaps if they had not been mismanaged and corrupt, they might have managed to pivot and find a profitable way forwards, but "staying the course" was business suicide.
Ergonomics. In the 1990s we had a choice of GUI systems on Windows and a system using Telnet session on SCO Unix, yes THAT unix. Anyway we were able to use a THIRD of the staff to achieve the same output and input of medical and accounting data for sone 300,000 lives once we gave up on the GUI systems. Once learnt a few quickeys, we confiscated the mice, the system was SO slick. Clinical systems have gone backwards as they insist we stare at the screen all the time. A touch typist and the function keys never need do so. It would peeb with an error or wrong box.
What GUI/Windows does is gives the impression of being intuitive, but you can never speed up. It makes processing painfully slow. Clinical systems in the UK have never got back the speed and elegance of the disciplined TELNET/SSH sessions of old. Medical IT has gone backwards.
Gerry
If you want to sell lots of servers, selling clients will help to make sure that happens, because then people buy in to your software and hardware ecosystem. I managed a couple of hundred Sun sparc servers and a few thousand Sun sparc clients in the 90's and 00's. The customers needed a rock solid development environment that was similar to where the products would be deployed. Big projects, lots of developers, lots of testers, and so much documentation and as projects neared their end the numbers of meetings increased exponentially. People squeezed work in between meetings (during if it was a teleconference) to have an update for the next scheduled meeting that day. It was in interesting way to keep people updated and projects on schedule.
I see that as why, in my mind anyhow, it has taken so long for ARM to end up in the server space, there were no proper ARM development clients. And I know people will say you can login to remote servers or use emulators, but there is real feedback to programmers who can probe out the strength and weakness of hardware. e.g. A thing one developer noticed was that by simply rearranging the paths of the LD_LIBRARY_PATH environmental variable was that they could boost compile times by 50%. Another developer noticed that two SunBlade 1500's (hers and a colleague in the same office, that were both delivered on the same day), compiled at amazingly different speeds - one finished a build over night, and the other was still going until lunchtime (and both had full duplex gigabit). In the end, once everything else was tested and eliminated, it was eventually traced back to Sun increasing the L2 cache in later revisions of the CPU used.
support tens of thousands of users, in as little memory as a first-gen smartwatch
"But you try and tell the young people today that... and they won't believe ya'." We were so poor
CDE was actually far more capable than most people thought. Not only was it a window manager and application launcher on your local system, it was also an authentication manager and a distributed application framework, which allowed you to treat a network of systems more as a single system.
While a lot of system admin's I know used to use rsh and then ssh as rsh fell out of favour to run commands and start up sessions across the network from their window manager of choice, CDE had this all built in, and IIRC, could use external security frameworks like Kerberos to manage authority.
You could create a task that you could trigger through CDE, and it would check that you had the authority, issue the command, and handle various I/O data flows between the machines, all from inside the CDE framework itself. I did not get much further than this, but I remember that there was more functionality than I ended up playing with.
It was not revolutionary, because as I said, these things were being done before, but it was all in one place. But it was a bit of a pain to set up (although it could have just been different I suppose) and use, IIRC.
Perhaps the greatest benefit of having multiple versions of UNIX and hardware that it ran on, was improving the quality of open source code. Porting software from one platform to others was a great way of finding glossed over bugs due to the differences in software and hardware. The downside was having to deal with the different bugs in the hardware and software. This is why the Open BSD group likes to develop on multiple hardware platforms as a software error may only be evident on one platform, e.g. the YACC bug found when running on SPARC.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
