▲ 204 points
back
34 comments
60's and 70's IBM was a tour de force. They're one of the first companies to realize that Design is an integral part of business and Design is not decoration. This was a top-down push, IBM was blessed by Eliot Noyes [1] the executive who had an architecture degree from Harvard Graduate School of Design. IBM proved to the world that everything from corporate logo (designed by Paul Rand) to their product design, stationary, building design, ... everything needs to feel uniform and part of the same company.
IBM System 360 was a pinnacle of design, brainchild of Eliot Noyes along with many other products from IBM.
Fun seeing that Model 20. The first commercial web site I ever built was in 1995. When I'd go up front to check the office fax machine or get some printer paper, I'd see the accounting department's S/360 sitting there, and wonder if I'd ever get to work on a "real" computer like that one.
Of course, by '95 the Model 20 was an ancient piece of junk, but it looked like serious business compared to what I was doing goofing off with this "world wide web" thing.
Five nines on a single CPU is hardly junk ;-)
But in 95, I'd guess it should be a 370.
Somewhat related, and mentioned here before, the project to recreate a functional scaled down front panel of the PDP-11 was really impressive.
Image: https://static.wixstatic.com/media/ce503a_7fce2e8386554dd1b8...
Project: https://obsolescence.wixsite.com/obsolescence/pidp-11
Mobile chrome has an annoying behaviour where it downloads all .webp image files via direct link instead of opening them as images in the browser. Seems to be an odd choice considering chrome’s otherwise support for them.
I removed everything after the .jpg and it fixed it. https://static.wixstatic.com/media/ce503a_7fce2e8386554dd1b8...
Thanks, I just wanted to complain
If you have the chance, take a look at the assembly for the 360/370. It actually made a lot more sense than the 8086. When I was in college, the Assembly class was on a 370 and I did love the documentation including the banana book ( https://www.flickr.com/photos/dabcanboulet/131534936 ).
As of about 10 years ago they still made those reference cards. And you're right about it making more sense than the 8086 if others are interested what you want to look for is the Principle of Operations document. I could only find the one for z9[1] but z13 is the latest mainframe out and there should be some somewhere.
Another interesting thing is that while you can technically run a version of Linux on these mainframe, most of them run z/OS with Unix System Services. So it acts somewhat familiar but you're sometimes forced to boot up a 3270 terminal emulator to connect and do some non-unix things.
[1] https://www.ibm.com/support/libraryserver/download/dz9zr006....
The latest one is for z14 [1] and is a very well organized document for compiler writers. I reference it daily while working on the Eclipse OpenJ9 project. Indeed from day to day we mainly work on Linux (Ubuntu 18.04) on our z14 mainframes.
[1] https://www-01.ibm.com/support/docview.wss?uid=isg2b9de5f05a...
Oh interesting, that's definitely changed since I was writing compilers for Z. Neat!
360 assembly is interesting because it's mostly normal instructions with a nice orthogonal register set and simple memory addressing. But then it has weird complex instructions for historical reasons. E.g. copying the top two bits from each character in an arbitrary-length field, or doing arithmetic on arbitrary-length BCD with two different formats. There's also the Edit instruction, sort of like a printf instruction to format numbers with commas. Overall, the instruction set seems like what you'd get if RISC guys collaborated with the worst people from x86.
It’s shocking to see serviceability and debugability as such a central part of the design and facade. These days most companies seem institutionally incapable of comprehending that failure is even possible. Customer support only exists to tell you how you’re using it wrong. Compare front and center observability of internal state vs. today’s hermetically sealed “just works” platforms and it’s clear that a monumental shift has taken place. I’m grateful that modern CPUs still support breakpoints.
It’s not “these days”, that problem has been with humans since the start of civilisation. Douglas Adams has a fantastic quote about it:
The major difference between a thing that might go wrong and a thing that cannot possibly go wrong is that when a thing that cannot possibly go wrong goes wrong it usually turns out to be impossible to get at and repair.
this is not user serviceability -- have you noticed phrase like
> supporting system maintenance and repair performed by an IBM customer engineer
Even today, if you are a service engineer servicing expensive piece of equipment, you have some amazing tools. I know an engineer from NetApp for example, and he has most amazing debugging stories, like debugging customers' systems live to find out where the latency spikes come from. The main difference from IBM era is that in the old times, customer could watch the blinkenlights, and now it is all hidden under password-protected accounts.
In just 55 years, this is like looking at alien technology compared to what we use today. Humans are notorious for looking at the incredible advancements of technology in the previous few decades and then saying, "and now we're done".
The idea that what we have today could seem as ancient and antiquated as what we had 55 years ago seems now; it is truly hard to fathom.
Maybe it won't? I think it's unrealistic to assume that computer technology will continue to see the kind of exponential growth it has seen in the last 50 or so years. Of course, I don't have hard evidence to back up my claim but I feel like most processes exhibit diminishing returns after a while.
Go on Youtube and watch 'The Computer Chronicles', especially the very old episodes. You can see that many people thought the same thing 30-40-odd year ago. Many experts and CEO's and visionaries on the show say things like 'that will never be feasible' for this and that reason. Smart people on the forefront of computer tech at that time.
I do feel the same way as you do though, but I am from the 70s and I am very careful in saying things like that (like you are by the way).
It's interesting that the people are wearing short sleeves, because it was usually freezing in the computer room. Also deafening. The amount of air conditioning needed to cool the computers also resulted in keeping it fairly cold in the room.
Had me curious, so I found the installation manuals. Apparently they recommend 75F, so perhaps not too cold: http://www.bitsavers.org/pdf/ibm/360/fe/GC22-6820-12_System_...
Though I've certainly been in some cold raised floor environments.
Thinking about it more, one possibility is that they kept rolling in more and more efficient machines, but never updated the HVAC system. Also, if the floor isn't well sealed, i.e., missing tiles, the room will be colder than necessary.
75F is cold if there is a strong breeze blowing on you all the time. These rooms had lots of airflow.
It was chilly but not so much freezing.
But those short sleeved people were clearly not IBM engineers.
I wonder if someone can build a web-based "simulated" version of one of those 360 mainframe consoles so we can play around with the switches and lights and get the feeling of it.
I wish I could!
I'm actually working on a web-based Model 50 right now. Bitsavers has most of the necessary documentation, but if anyone has the ALDs (automated logic diagrams) let me know.
I was surprise to see the meter. I thought they use punch card to measure usage (and have a major antitrust case on this).
For the comment, it never was that cold except when I have to install the machine OS, when other part has not been installed yet and tile is open. Noisy. Very noisy though.
*
Wish there were more history though. That is too long ago.
First use of IBM is already 4381. By the time I Am the system administrator of one, it was using LPAR and ESCON I believe. Can’t remember the series. So enjoy then the assembler can have more than 24 bits. ESA I believe. But the mainframe people attitude is hard to get around to (even though I should be). I bought a pocket PC to calc (hp dos) and was attacked by everything like i being evil to the heaven.
What kind of "business reports" did these things process? I'm curious what application for your typical business was high value enough to be worth all this complexity and cost.
Businesses used punch cards for accounting, payroll, inventory, taxes, billing, and lots of similar applications. Way before computers, back to the early 1900s, businesses used electromechanical accounting machine to process punch card data. By 1944, there were over 10,000 of these accounting machines in use. Low-end computers often were a replacement for these accounting machines. The language RPG (Report Program Generator) was often used to generate business reports.
There are lessons to learn from it. The dials, knobs, sliders, the whole shape of the thing -- the gravitas, speaks oodles of cash and it looks well worth it to a CIO. These days we have web pages. A web page with right look and feel can mint money too.
Duplicate discussion at https://news.ycombinator.com/item?id=19598389 .
I hope that lady with the metal bracelets was careful that they didn't slide forward and short out something.
It would have been a serious design failure to have shortable parts on the operator front panel.
I'm sure there would have been a man close by to remind her about basic parts of her job constantly
Wow, that IBM System/360 Model 85 is sexy af.