Come for ENIAC, Stay for UNIVAC and Skeduflo | Hacker News Reader

Come for ENIAC, Stay for UNIVAC and Skeduflo

uniqueatpenn.wordpress.com

60 pointscainxinth28 comments

Google had never heard of Skeduflo. I hadn't either, and I once did work on critical path programs using UNIVAC computers. It was apparently an analog computer for computing PERT charts. It probably never got beyond the prototype, or there would be more history around it.

That's a neat idea. I can see how it would work. A PERT chart is a time-oriented acyclic dependency graph. Each task has a duration, and tasks can be in parallel. At join points in the graph, the later end time determines the time value for the end node, and that feeds the following tasks.

So an analog computer could do this with voltage=time and resistance=duration. The picture shows a compact plugboard at the lower right. That's the dependency graph. Then there are lots of knobs, for setting task duration. For each join point, there would be a circuit to pass through the higher voltage (maybe just diodes, although diode drop would be a problem; might need an op amp). Then you have lights to tell you which path is longer at each join point. Now you can tweak the knobs and adjust the schedule. There ought to be a meter available for reading out times, but maybe that's external.

My grandfather started Mauchly Associates with John Mauchly, he was Mauchly's "business guy" because he had experience in computers while working at Remington Rand on the Critical Path Method[1] in the 50s. My mom remembers John telling her about someday they're going to make computers that can fit in a briefcase while the smallest computers at the time were the UNIVAC. Not sure why you don't hear much about Mauchly Associates, it's not even on his Wikipedia page...

[1]: https://en.wikipedia.org/wiki/Critical_path_method

Great story, thanks. I wasn't aware that the Critical Path Method originated from DuPont and Remington Rand. Almost lost art today. When I proposed it in a government project twenty years ago here in Switzerland nobody seemed to have ever heard of it.

I was hoping Minnesota would be mentioned, I had an older relative that once worked for UNIVAC. I didn't realize how spread out it was.

From Wikipedia:

>The UNIVAC was manufactured at Remington Rand's former Eckert-Mauchly Division plant on W Allegheny Avenue in Philadelphia, Pennsylvania.[1][9] Remington Rand also had an engineering research lab in Norwalk, Connecticut, and later bought Engineering Research Associates (ERA) in St. Paul, Minnesota. In 1953 or 1954 Remington Rand merged their Norwalk tabulating machine division, the ERA "scientific" computer division, and the UNIVAC "business" computer division into a single division under the UNIVAC name.

I'm guessing all the coolest stuff had long since happened before my relative was there.

I worked at the Roseville, MN, plant in the early 80's as a CPU designer. 100K ECL. The machine was the Integrated Scientific Processor, an attached processor that could be added to an 1100/80 system. Max configuration was 4 CPU's (model 1100/84) plus 2 ISPs. The ISP was a vector processor that had approximately the same peak performance of a CRAY-1, and higher average performance. (In part because we handled the "short vector problem" more elegantly, so the last few elements of every long vector didn't pay a time penalty for dealing with the tail end of a long vector. There were a few other architectural tweaks that contributed.)

The program was not a success. Reason 1: Oil prices came way down, so seismic data reduction and reservoir simulation markets dried up, Reason 2: Univac management was timid and slow at decision-making and starved the project. Had we been properly resourced and delivered a couple years earlier, things could have been different.

Anyway, it was a hugely fun CPU to work on.

This article has a striking absence, that being the name "John Atanasoff"[0], whose electronic digital calculator was seen and examined by Mauchly in 1941, and who spoke to Mauchly multiple times during the years when ENIAC was being built.

It was this connection that ultimately blocked Eckert and Mauchly from patenting the idea of the digital computer.[1]

[0] https://en.wikipedia.org/wiki/John_Vincent_Atanasoff [1] https://en.wikipedia.org/wiki/John_Vincent_Atanasoff#Patent_...

I heartily recommend both (if you can find them) "The First Electronic Computer: The Atanasoff Story" (for a technical look at the computer itself, by Arthur and Alice Burks) and "Who Invented The Computer?" (for a history of the ABC as it applies to the patent trial, by Alice Burks).

> It was this connection that ultimately blocked Eckert and Mauchly from patenting the idea of the digital computer.

Why would a connection be relevant? A patent can be blocked by prior art, but that theoretically happens whether you know about the prior art or not.

Knowledge is relevant to how guilty you are over infringing an existing patent, and it's relevant to copyright, but I didn't think it was relevant to patentability.

…and whose digital calculator was _just_ a calculator (not a general purpose computer) that read values from a mechanical drum (so not fully electronic), and sat in a basement until it was later disassembled. That this device managed to qualify as prior art is mostly just an exhibit of how the trial system isn’t especially good at divining truth. So annoying that the ABC is a permanent barnacle on the ENIAC story.

The ABC's memory was not mechanical, even if the drum was driven directly by an electric motor. It was a self-refreshing series of capacitors using the same general mechanism as modern, non-persistent memory (self-regenerating I think?). It's was one of the other parts of the ABC that invalidated the Eniac patents, in addition to the ABC's binary adders. But the memory itself was entirely digital.

Sorry, I shortened to the point of inaccuracy - I should have said not fully electronic _speed_. Not meant to say that it was mechanically reading and writing values per se, just that the slow rotation of a mechanical drum was essential for its operation and ultimately had a limiting effect on the speeds it could compute. Thus the electronic storage and compute was rather defeated by inclusion of a mechanical element in the “critical path”, so to speak. I think that’s accurate to say?

Not any more accurate than it is to say the electronic storage and compute of any computer is defeated by the latency of its memory. Virtually all early memory devices were deeply limited.

Really the ABC was most limited by its use of a custom I/O device which used electrically burned holes in special paper to input the initial problem state and output the solution. Atanasoff never got it working reliably, which is the main reason the ABC was never fully functional before WWII.

In any case, the fixed nature of the ABC is often why its considered the first "digital electronic computer" and the ENIAC the first "general purpose digital electronic digital computer" (or later, the first "programmable digital electronic digital computer", after it was rebuilt in '46 - '47 to operate more like a modern computer).

Unfortunately I'm not familiar enough with Zuse's computers or the Colossus to say if those statements are really accurate.

Quoting a Time article in 1962[0]:

> Definitely not yet on the market, but envisioned by Dr. John W. Mauchly, is a miniature computer for household use that will not only make shopping lists obsolete but will also mark the extinction of the grocery clerk and the checkout-counter man. Before going to market, a woman will slip her computer into her purse (it will have an inventory of what she needs in the way of staples and supplies stored in its wafer-thin memory cells). Once at the market, she will plug her computer into a socket in a vacant “delivery alcove” and wait for the results. The computer will carry out the business of identifying itself, making the proper accounting entries in its own memory, and authorizing the charge against its mistress’ universal checking account. In less than a minute the order slides down a chute, and the housewife brings home the electronic bacon. Dr. Mauchly, who invented some of the original big computers and has already built one the size of a suitcase, is working on the pocket monster.

[0] https://time.com/archive/6625605/the-marketplace-build-small...

Today's surviving UNIVAC architecture originated in the first transistor-based machine, the 1107, which was not compatible with any of the previous designs that all used vacuum tubes. "Only 36 systems were sold."

https://en.wikipedia.org/wiki/UNIVAC_1100/2200_series

The 1108-II was the first SMP dual processor architecture that was sold commercially, announced in 1965.

The operating system launched on the 1107 is now known as OS 2200, and is still supported.

https://en.wikipedia.org/wiki/OS_2200

United Airlines remains a major customer for the UNIVAC architecture.

The first SMP Unix implementation was on UNIVAC hardware, and used the EXEC-8 kernel from OS2200 directly, implementing a Unix system call as a "shim" around it.

https://www.nokia.com/bell-labs/about/dennis-m-ritchie/other...

"The UNIX system for the UNIVAC 1100 series was built as an integrated development environment for transactions that run directly on EXEC. Unlike most other implementations, therefore, it runs not directly on the hardware but as a collection of user-level activities under control of EXEC. These obtain services that would normally be provided by device drivers, and some process creation and management services from EXEC. Any configuration supplied by Sperry, including multiprocessor ones, can run the UNIX system."

For a while I worked at what was then the Sperry Rand Corporation (now Unisys) which had some pride in their heritage as the descendant of the Univac Corporation founded by ENIAC inventors Eckert and Mauchly. In a glass case there was a vacuum tube circuit said to be a memory unit of the original ENIAC. No one seemed to know much about it, casting doubt on the claimed provenance of the device. The tube circuit resembled the ones shown in the photo linked below (although none of those in the photo are from ENIAC).

https://en.wikipedia.org/wiki/File:Women_holding_parts_of_th...

(This is a repost of an older comment I made)

If it looked like the one on the left then it was probably an ENIAC decade ring counter. It could store 1 digit from 0 to 9 and could selectively count pulses to increment the value, with carry to the next one. The 10 tubes on the right (out of 36) were the actual flip flops of the counter.

I recently read a fascinating book I found in the library by an engineer who worked under Mauchly and Eckert and talks about working on ENIAC, BINAC, EDVAC, LARC, and UNIVAC. The book is "From Dits to Bits" by Herman Lukoff.

That’s a really good book!

Yes! Thanks for recommending that so I don't have to...

It's also available on archive.org. Told by a guy who was there for a lot of the formative stuff.

I worked for Sperry Univac 1974/78. Great bunch of people. It’s a shame their resources were diluted over so many different operating systems.

Wow. Must feel like a completely different world compared to today. May I ask what your work was?

I was a Systems Analyst. I helped salesmen prepare proposals, I did benchmarks and conversions, and customer training. Univac couldn't afford to put a computer in every branch, so I had less access and opportunity to program, than our customers. Made support awkward.

We also had internal competition and friction. The 9000 and 1100 teams resented if their technology was not pitched to a prospective customer. The 9000 was lower-end than 1100, but there was enough overlap in the mid-range.

I also programmed the CADE 1900, a data entry system that paid for many trips throughout Canada and the U.S.

Before that, I wrote apps in IBM mainframe Assembler for a life insurance company.

(Sorry I didn't reply sooner, I had not seen the question).

Thanks for these interesting insights. It's hard to imagine today how scarce computers actually were at that time. The assembler was for the System/360?

The Assembler (Basic Assembler Language or BAL) was for the Univac 9300 and 9400, clones of the IBM 360. Univac originally made and sold the 1100 series, with the Exec 8 O/S. The 1100 was a 36-bit machine, very successful in airline reservation systems, petroleum exploration, etc. but it was incompatible with IBM's 360 architecture. Univac developed the 9000 so IBM customers could easily convert.

The 9300 I programmed had 32K, tape drives, card reader & punch. It ran NCOS (Non-Concurrent OS). It had no virtual memory, so we had to write our own "overlays", sections of code that were loaded at beginning and end of a run, for initialization and wrapping up. No console, just sixteen lights to display error numbers, and eight butterfly switches for writing directly to memory.

Imagine running an insurance company on 32K!

That's fascinating. I spent significant time with early MUMPS and the machines of the time (https://github.com/rochus-keller/mumps/) which had a comparable amount of memory.

Was your whole insurance application only assembler, i.e. no COBOL or RPG at all?

The company started out with RPG. There was no COBOL compiler. They had to start processing health claims on paper tape, which could only be read with Assembler. They hired me to learn & write Assembler. We converted the master file update (40,000 cards, two hours twice a day) to Assembler, when I discovered a way to store 240 digits per card: twelve bits/column = three decimal digits. It was the 70's (7 = 0111), so our cards looked like lace doilies. Reconstructing jammed cards was fun.

I left the company after three years because I could only compile & test at night or on weekends.

Cool. So at least the work-hours haven't changed much since ;-)

The capitalization of "SkeduFlo" in the photo of the newspaper article made the intended pronunciation more apparent to me.

Reply on news.ycombinator.com