Advent of Computing - Episode 44 - ENIAC, Part II

Episode Date: November 30, 2020

In 1946 John Eckert and John Mauchly left the Moore School, patented ENIAC, and founded a company. One of those discussions would have consequences that wouldn't be resolved until 1973. Today we close out our series on ENIAC with a look at the legal battle it spawned, and how it put ownership over the rights to basic digital technology on trial. Along the way we talk legal gobbledygook, conspiracy, and take a look at some of the earliest electronic computers. Like the show? Then why not head over and support me on Patreon. Perks include early access to future episodes, and bonus content: https://www.patreon.com/adventofcomputing

Say what you will about the free market, but competition has been a massive boon for computers. I know I run back to IBM probably a little too often, but I'm going to do it again here. The race to the bottom of the PC clone market in the 1980s is a perfect example of how competition made computing better. IBM's original beige box was relatively expensive. But soon after release, other manufacturers figured out how to make compatible systems. And what's better, the compatibles came at a fraction of the price. As the PC market opened up, standards started to form.

Computers became more accessible to more people. And third-party hardware and software flowed freely. But at first, this market was a little bit tenuous. The key to making this all possible was that the companies behind PC clones found a safe path to building new computers. That's one that kept them safe from IBM's lawyers. With schematics for the PC widely available and cleanly reverse-engineered BIOS code, third-parties were able to skirt around lawsuits. cleanly reverse-engineered BIOS code, third parties were able to skirt around lawsuits. Thanks to that, the PC has become one of the most widely used platforms in the world.

A winning design was able to escape from the walled garden, and really, we've all benefited from that. However, compared to earlier events, the PC clone saga is pretty small potatoes. The PC clone saga is pretty small potatoes. Back in the 1950s, before squabbles between IBM and Compaq, a much more grand legal struggle played out. One that held the fate of commercial computing in the balance. The question at hand was, who owns the idea of a computer? Welcome back to Advent of Computing. I'm your host, Sean Haas, and this is episode 44, ENIAC, part 1. Today, we're going to wrap up our dive into the sordid past of one of the first computers.

Last episode, I tried to explain, at least as best as I could, why someone would ever want to build a computer in the first place. To do that, we took a look at ENIAC, the environment that led up to it, the machine's development, and then examined early uses of this computer. But here's the thing, that's not really the full extent of this rabbit hole. ENIAC itself actually had a fairly long and somewhat productive life. In 1956, it moved from the Moore School to a lab at the Aberdeen Proving Ground. After some upgrades, it continued to operate all the way until 1955. So for about a decade or so, people were still patching together programs on this downright archaic machine. But that's not what we're going to be talking about today. When ENIAC moved, its initial designers, John Mouchley and

John Presper Eckert, didn't move with it. Instead, they partnered up and founded a computer company. In fact, it was the first computer company. Now, that's all well and good, but when they left the Moore School, they planted the seeds for a pretty strange happening down the line. In 1946, the team of Johns filed a patent for ENIAC. A lot of litigation and near disaster would stem from this single decision. By the 1960s, the patent would spawn a legal battle, one that would be longer than any other case in U.S. history, well, at least up to that date. What makes this all the more interesting, and the reason I wanted to cover it, is who winds up being involved in the patent disputes and ensuing trial. It turns into a real who's who of early computing. And when all is said and done, the case law established will make competition

in computing possible. For our purposes, the ENIAC Patent Saga holds a lot of answers. With so many people involved, we can start to see why computing mattered back in the very early days of the field. So let's hop in. This episode, we're going to run into some old friends, a few new faces, and see how one of the first computers affected the industry from far beyond the grave. If you humor me here for a moment, I have a bit of a pet theory. I personally think that the ENIAC team never slept a full night in their lives. If they did, it must have been during some stolen moments in nearby supply cupboards. As early as 1944,

before ENIAC was even completed, John Mauchly and John Eckert were already working up plans for a new computer. This second system, EDVAC, or the Electronic Discrete Variable Automatic Computer, was planned as a massive improvement over ENIAC. The most consequential change was that EDVAC would be a stored program computer. Last episode, I dedicated a whole lot of time to discussing how ENIAC was programmed. But to quickly summarize, it was a bit of an ordeal. ENIAC didn't have an instruction set. That meant that you couldn't write code for the machine. Instead, it was built of separate modules that had to be wired together in sequence to build up a program. Now, there's

some pros and cons to this arrangement. The pro is that it made ENIAC a lot quicker to develop and build. The cons are basically everything else. It made programming ENIAC a monumental task. basically everything else. It made programming ENIAC a monumental task. It was a known problem, but with the time crunch on the project, some corners just had to be cut. But hey, once the war was over, there was no time crunch anymore, so the Johns started making plans for peacetime a little early. Those dreams of a new computer floated around the Moore School, and as ENIAC was coming to a close, the team shifted to start designing EDVAC. Once Los Alamos got involved with ENIAC, John von Neumann would also join the EDVAC team. By 1945, an all-star cast had really formed, and so too had a concrete design.

And yes, if you're keeping track at home, we're back up to three Johns that matter. The basic plans for EDVAC were completed in early 45, and by June, a report was drafted up by John von Neumann. The timeline here is what I think adds a little bit of credibility to my theory. Before ENIAC was ever running, the team already had grander ambitions. The report, titled The First Draft of a Report on EDVAC, has all the gritty details of those ambitions. But what's interesting is that within the hundred or so pages, it describes a machine that's totally different from ENIAC. Without time restrictions, there was no reason to make compromises. The result is a computer that's

much more in line with modern systems. One of the little nitpicky things about the report on EDVAC that I have to mention is the actual proposed circuit design. Unlike ENIAC, there is a least common denominator to all operations. Von Neumann called them E-elements for... some reason that I can't understand. But the smallest operation on EDVAC was actually made up of simple logic circuits. Each operation, say an AND or an XOR, can be built up using vacuum tubes. Then from there, more complex things like math operations can be made from these simple building blocks. The report even explains how multiplication circuits can be built using logic gates.

It may sound like a small change, but it speaks to a lineage starting to form. Modern computers, or really any computers built after ENIAC, adhere to this design philosophy. You don't have custom one-off circuits that run special operations. Everything is built using chains of simple binary logic. The other big implication here is that EDVAC was a fully binary machine. It didn't have fancy registers for storing base 10 numbers. Everything was either a

1 or a 0. You're on or you're off. While you can run decimal operations using digital circuits, it's not a natural fit. But binary goes with digital like peanut butter and jelly. Binary numbers form a perfect representation of digital circuitry. It's a matter of using the right medium for the right message. And in that, EDVAC was taking a step in the exact correct direction. We also start getting to a point where we can pretty clearly identify one-to-one components between EDVAC and computers that we currently use. All math operations were built into a single module called the computation unit. In more modern times, we call that an arithmetic logic unit. A control unit was built for managing instructions, which is still present by the same name in modern processors.

The laundry list is already getting more in line with what we expect inside a big box. But as far as I'm concerned, here's the biggie. EDVAC also had memory. And I'm not talking about the weird function tables and accumulators that ENIAC ran with. I'm talking full-on accessible, mostly random access memory. The final implementation was a bit strange. It eventually used Mercury Delay Lines, which store data as these bouncing pulses. But details aside, it's still memory. Constants could be stored in EDVAC's RAM,

so could changing variables. Most importantly, the computer's memory could also be used to hold programs. That last part I really want to stress because that's key here. EDVAC was designed to run a series of commands, and those commands were stored in the same memory space as data. It all just goes in the big tank full of mercury. This is how most computers work nowadays, and there's a good reason for it. The most readily apparent reason, at least on these older systems, is that storing code and data together saves you some money. Instead of having two memory devices, you can just have one. Treating code and data the same also lets you pull some pretty sick tricks. You can modify

a program while it's running, which gives you a little bit more flexibility. Or, on a more mundane level, you can load in programs from long-term storage, such as punch cards or magnetic tape. load in programs from long-term storage such as punch cards or magnetic tape. A program can be kept around indefinitely, and when needed, you just read it into the computer's memory like any other data. This whole arrangement, plus more technical parts like bus architecture that made it all possible, is so important that it actually has a name. It's become known as the von Neumann Architecture.

The draft report on EDVAC would prove to be immensely influential. Not only did it help the team secure another military contract, but it spread widely outside the Moore School. Von Neumann's draft report pretty much explained how to build an electronic digital computer. It came out in a period where the first computers were just starting to run, so a lot of people took interest in it. If you want to sound a little grandiose, then it was really the first piercing of the veil around computing technology. Systems like ENIAC were still pretty secretive. At least their internals were. But the EDVAC report? That laid out everything. If you had enough funding and the right people around, then you could build a computer. That sounds fantastic, right? The

EDVAC report was a spark that would go a long way in the development of the discipline. But things are a little complicated here. The cover sheet for the report is pretty standard. It has the paper's title, some information about who was funding it, and a date. It also has a byline, and it's a noticeably small one. It just says it's written by John von Neumann. If you were one of the many researchers in the 1940s that read the EDVAC report, then you'd think this brilliant computer design was solely created by good old Johnny von Neumann. It should be clear, though, that nothing could be further from the truth. Von Neumann consulted on the design, he put everything to paper,

and he made the report more presentable in an academic form. But he wasn't working alone. In fact, he wasn't even the main contributor. The ENIAC team did the majority of the actual work, and according to Eckert and Mouchley, EDVAC was their idea. They felt that this third John just came in and stole all the credit. It must have added a little bit of an extra sting when the design started being called the von Neumann architecture. Maybe a better word would really be the John, John, and John architecture.

I think Jean Bartik, one of ENIAC's programmers, really put it best when she said this in an interview. Quote, well heck, von Neumann got all this stuff from present John from the EDVAC because he did not invent the stored program computer. And in fact, this business with the EDVAC, I mean to say that everybody knew they had IOs and different things to do. The thing he did was to write it down, end quote. The he there is, of course, von Neumann, the one who finally devoted things to paper. Now, there's actually some weird contention about how the report ended up being published in the first place. It is a draft after all, it's not the final report. The most commonly repeated story that I've seen is that the report was intended to be internal

only, and that Herman Goldstein, one of ENIAC's main engineers, sent out copies of the document. According to Bartek, when confronted, he claimed that he didn't intend for it to be published, so he didn't bother including proper credits on the draft. There are also some allegations that Goldstein may have been trying to steal credit for EDVAC, but we get into some territory where we only have interviews about 50-year-old arguments to go off of. I didn't dive too deep into that since it's a little tangential to the main story, but there were definitely some strange dealings at play here. Credit wasn't the only problem. They had another reason to resent what had been done to their work. For a while, Eckert and Mouchley had been talking about leaving

the Moore School and going into the private sector. The contracts that they were making with the military were really lucrative, and they figured, rightly so, that there could be a much larger market for computers. They had intended to patent ENIAC and EDVAC, and then go into business using those patents as the cornerstone of their company. But with EDVAC's design now out in the open, they could no longer patent it. Possible treachery was just one of the problems that was facing Eckert and Mouchley. The other big problem on their mind came from the Moore School directly. The policy prior to 1946 was that professors and researchers retained patent rights for anything they created at the university. But, you see, that's not the most lucrative arrangement when the college is concerned.

It leaves a lot of money on the table, and administrators don't like missing out on dividends. So it was decided to change that policy, so that any patent filed by employees would be owned wholly by the Moore School. For Eckert and Mauchly, that was the last straw. The two quit, and they went off to form their own computer company. In 1946, the duo founded the Eckert-Mauchly Computer Corporation. Now, I know, that's a pretty inspiring name. The company, EMCC for short, was the first private sector computer manufacturer in the world. Up to this point, computing had been firmly in the realm of public institutions and some large labs. Once again, Eckert and Mautley were treading

totally new ground, but they knew what direction to go in. Their initial business plan states the purpose of the company pretty well. Quote, the first objective of the proposed company will be to design and develop multi-purpose rapid computing machines of moderate cost. End quote. In no uncertain terms, the plan describes how the experience gained from ENIAC would be used to make better computers. Think of it as a splinter project to rival EDVAC. The other interesting part of this early plan is that EMCC wasn't only going to target military contracts. Instead, the document paints a slightly more egalitarian view of computing. Well, it's egalitarian if you're a business. Eckert and Mouchley saw computers being usable for bulk data processing,

listing insurance companies, the Census Bureau, factories, and laboratories as possible clients. So the idea of who would benefit from a mess of wires is already starting to expand a little. But the core of this plan, and what everything would ride on, was locking down that patent. That process would get started in 1947. locking down that patent. That process would get started in 1947. That's when the duo of Johns actually submitted the paperwork for a patent on ENIAC. This is where we take a turn from pretty academic fights into more dangerous territory. No matter which way you slice it, ENIAC's patent, number 3120606 is massive. It spans over 200 pages, half devoted to figures and diagrams and half to actual text. If you've spent time looking at patents before, then you know that those numbers are a little bit out of the ordinary.

But it's not just long. This patent is also really broad. Just to give you a taste, here's the opening paragraph. Quote, and more particularly relates to the art of electrical computing machines with particular reference to a machine using electronically produced pulses, i.e. sharp voltage changes not greater than 5 microseconds in duration, to represent digits and numbers and using such pulses for control and programming operations, thus obviating the need for mechanically moving parts for these purposes. That's all one sentence.

And it's awful. The entire patent is basically written like that. So, you know, if you want to go read it, maybe just don't. Anyway, if you don't understand patent office English, then what exactly are Eckert and Mousley patenting here? Simply put, the entire idea of an electronic digital computer. The binary pulses mean binary, or rather the electronic representation of binary data. But that's just the start of the document. Patents are usually laid out something like this.

You have your diagrams of the invention, a long chunk of text that describes the patent and diagrams, and then a list of legal claims to ownership of certain ideas. That last stuff, the enumerated claims, are where we actually need to focus, at least for the upcoming legal stuff. And as with everything about ENIAC, this section is huge. In all, there are 148 claims. What's key here is that Eckert and Mouchley are claiming 148 features of ENIAC's design that, according to them, are totally new and totally unprecedented. As in, no one has previously came up with these

ideas. But I'll level with you here. This is one of the most boring and tedious documents that I've read to the show, at least so far. I don't know if it's just me, but, you know, patents have this wonderfully painful quality to them. It takes all the structure and repetition of a legal document and adds in all the terse and confusing language of a technical paper. That being said, here's one of the claims that I think is very important. That's claim number nine. It goes something like this.

Quote, An electronic computing machine comprising arithmetic units of respective numerating function. It goes something like this, quote, such pulses, and also being responsive to individually characterized control signals to perform respective arithmetic operations on numerical data so transmitted thereto, means included in each said unit to emit pulse signals significant to the numerical results of respective arithmetic operations completed theretoin, and means included in each of certain at least of said units responsive to completion of repetitive arithmetic operations in said units to emit one at least of said control signals, characterized significantly in relation to the integration of results from two or more said units. End quote. Once again, that's all one sentence. Take this as a sample and save yourself the struggle of reading the patent.

So, in more reasonable terms, what are Eckert and Mouchley actually laying claim to? Well, claim number nine describes the core functioning of ENIAC. You have a series of modules that each perform some distinct mathematical operation. The manner in which data is shuttled around, in this case digital pulses, on a wire. The mathematical meaning of the data in those pulses, and finally, all the hardware that makes this function. Further on, the patent uses pretty similar language to lay claim to logic and control operations. Just as important is what these claims don't specify. ENIAC is a strange computer. If you listened to last episode, then you should be pretty familiar with all the details that I mean when I say that. A lot of those idiosyncrasies don't show up in the patent at all. Taking claim

number nine as our example, it never describes what kind of data is being sent around the machine. It also doesn't describe which mathematical operations or how data is stored. Just that math happens and data moves around as digital pulses. What do those pulses travel down? How's the data actually encoded? That's not relevant here. The patent doesn't just describe a general purpose computer. It describes in very, very general terms means to transmit their two numerical data and distinctly characterized control signals and pulse form. Well, that can be applied to basically any digital device. It's not unique to ENIAC. The main reason to list claims in such general terms comes down to patent law. Eckert and Mauchly weren't just going to patent ENIAC and then rest on their

laurels. The grand idea was that they were patenting a collection of technologies. This would allow EMCC to corner the market on computing in general, not just on clones of ENIAC. I also think that the Johns took this tact as a direct result of the whole EDVAC debacle. Much of the technology in the EDVAC report was similar, at least in general terms, to technology used in ENIAC. Eckert and Mouchley couldn't patent EDVAC, but they could do the next best thing. By going upstream a little bit, they could lay claim to the underlying tech that made EDVAC work. Once again, here's where ENIAC's legacy becomes more conflicted. Is it ethical to patent something as broad as a digital computer? To us, or at least to me, the answer is no. But that opinion is very heavily influenced by the power of hindsight. It's a superpower that people in the 40s didn't have yet. From the 21st century,

it's really clear that computers have had a major impact on the public good. Access to digital technology has made the world better. It's helped countless people. And it's just expanded our horizon of knowledge. But in the 1940s, that wasn't really the intention. Remember that ENIAC grew out of America's war machine. It was developed as a tool for very non-altruistic purposes. To Eckert, Mauchly, and really everyone involved, patenting ENIAC must have been no different than patenting a better kind of bomb. Ethics aside, there were also more mundane complications for EMCC. The patent for ENIAC was filed officially in 1947, shortly after the first computer company was incorporated. But that's just the first step in getting a patent issued.

The application has to go through a long process of validation, review, and possible legal challenges. With something as large and important as ENIAC, this process became very extensive. Each claim has to be checked to ensure it is actually new. If an earlier patent, termed a patent with priority, makes the same claim, then you run into problems. And for a patent as large and important as ENIAC, this process took a whole lot of time. Hadn't as large and important as ENIAC, this process took a whole lot of time. The ongoing review became a background to EMCC's daily operations. On the purely business side, there were some other problems.

Despite having a reasonable business plan, some of the best minds in the field, and a brewing patent, EMCC had ongoing problems as a company. Their first project, UNIVAC, was designed over EMCC's first few years of life. And in 1948, they won their first contract, selling a UNIVAC to the U.S. Census Bureau for the 1950 census. It was a really promising start, but building and debugging the machine took longer than expected. EMCC would also win a handful of contracts for the US military and government, but things only went downhill from there. The same year contracts started flowing in, the company took a serious blow. And this is a bit of a weird one. A group of engineers, including John Mouchley himself, were accused by the FBI of being

communists. That's right, they got hit with McCarthy-era investigations. EMCC lost their security clearance, and that meant that they lost their ability to gain government contracts. For a number of years, Mouchley wasn't even allowed inside the company's office. This was firmly during one of the red scares, after all. This persecution couldn't have come at a worse time. It put EMCC further behind schedule, and it cut them off from their main source of new clients. The company would pick up some smaller contracts to try and stay afloat, but the writing was on the wall. In 1950, the company went up for sale.

Soon after, it was bought out by Remington Rand. Just five years later, Rand merged with Sperry, forming Sperry Rand. And with each of these mergers, a portion of the ENIAC staff came along. And most importantly, the still-pending patent changed hands. By 1955, the much more well-funded Sperry Rand was now in charge of any legal fights over one of the most important patents in decades. Now, with that out of the way, I think it's about time to step into the first courtroom. Even after buyouts, mergers, McCarthy probes, and new offices, the all-important patent was still going through review. Did I

mention that this is a very large document? Well, you see, the sheer size and complexity was really only one issue. The other was a now-obsolete legal practice known as an interference proceeding. But it has another name that I kind of prefer. It's called a patent priority contest. I said that each claim made in a patent has to be something new, and any claim that comes before it is said to have priority. This is also applied to patents, broadly speaking. A patent submitted earlier is said to have priority. If I walked down to the local patent office and tried to submit my very well-written patent for sliced bread, I'd be turned down since some other patent takes priority over my claims of cutting up a loaf. If I could make some new way to slice bread

and show that no prior patent laid claim to my novel method, then I could get past review. patent-laid claim to my novel method, then I could get past review. Then, patent in hand, I could start charging folk to slice bread with the new Sean method. A priority contest is a separate way to block patents. It's more of an active approach. This boils down to an idiosyncrasy of how the US patent system functioned, or at least functioned up to 2011 when laws were revised. My sliced bread scenario is a first-to-file system, that is, a patent system in which whoever submitted the proper forms first gets all the spoils. But historically, the US has operated under a first-to-invent system. This is where, if you can prove you invented something first, you can claim rights. You don't technically need a patent, or at least no identical claim is needed. The legal mechanism

to enforce these claims was the priority contest. It's a process in which an inventor can sue to stop the patent review process. Now, it's a little bit of mumbo-jumbo to me. Here's how it works in the real world. Eckert and Mauchly claimed that they invented sending data as digital pulses and operating on digital pulses of data. They filed a patent that claimed that in 1947. Turns out that not everyone agreed with that claim, and inventors of earlier systems started to come out of the woodwork. During this review process, in 1959, Bell Labs was the first to publicly disagree. The nice people over at Bell Labs then sent a group of lawyers to Sperry Rand to note their disagreement. The lawyers at Sperry Rand said that Bell was wrong, so a court battle ensued. Thus, the patent for ENIAC was further delayed. What's so interesting

about this phase of the patent battle is that it pulled at least one strange proto-computer out of seemingly nowhere. For us, that's a good thing, since proceedings started to shed light on otherwise obscure technology. AT&T's Bell Labs was the first of these fights, and I think it's a particularly important challenger. Bell wasn't claiming that they had priorities on the grounds of computing, or at least not in total. They were claiming that they had priority on digital data communications with a computer. So yeah, this is a pretty strange one.

So let me take a minute to explain. Bell Labs already had a computer since way back. Well, they kind of had a computer. In 1937, George Stibitz, a physicist working at Bell Labs, started to experiment with digital logic circuits. This started with really simple relay circuits that he threw together on his kitchen table. Over the course of 1937, he started to work out how to perform mathematical operations using simple relays. That is, Stibitz created relay circuits for adding and subtracting binary numbers. By 1938, his idle tinkering started to pick up a lot of steam. Researchers around Bell had been looking for a way to offload some of the manual work

involved in complex calculations. By complex here, I don't just mean complicated, I mean calculations that include imaginary numbers. This type of math gets pretty messy pretty quickly, and as a result, it tends to take a lot more time than normal arithmetic. So Stibitz offered a solution. He figured that it should be possible to scale up his kitchen table circuits into a fully realized computer. Before 1938 was over, Stibitz would put the finishing touches on the Model 1 complex calculator. This was a relay machine that was capable of automating away a lot of Bell's problems. Now, it wasn't a fully-fledged computer. It couldn't perform conditional branches or loops, but it could perform complex operations faster than most

people. And it shared a lot of ideological foundations with ENIAC. Both machines were built to make math more manageable. Both relied on fast-switching electric components. And, in at least one way, the complex calculator was miles ahead of ENIAC. Stibitz rendered everything in binary instead of decimal. Once a number was inside the machine, it was fully binary, at least until results had to be printed out. Over the ensuing years, the design would be expanded and developed. By 1944, a fully programmable system called the Model 5 was under construction. But things get a little more sticky from there.

In day-to-day operations, lab workers didn't walk up to a bank of switches and enter in an equation. Instead, the complex calculator was accessed over teletype. A few labs around Bell's New York City campus were outfitted with simple terminals, basically tricked out electric typewriters. From there, a researcher could type in their query. It would then be turned into a series of digital pulses and sent down a line to the complex calculator. Some relays would clack, an answer would be produced, then it would be sent back down the same line to be printed at the teletype. In 1940, there was even a demonstration to an audience. Stivitz showed off his machine by

linking up a teletype at Dartmouth College in New Hampshire to a far distant calculator back in New York. In other words, Bell had created a, quote, means to transmit thereto numerical data and distinctly characterize control signals in pulse signal formats. Bell's suit wasn't just over who made the first computer. It was also over who had the rights to digital communication. Bell held a patent simply titled Electronic Computer, essentially a more formalized version of Stibbitt's relay calculator. This patent sounds really close to ENIAC, save for the lack of programmability and some changes to language. Bell's claims even describe a, quote, chain of operations, where data travels via pulses on a wire.

But some issues start to show up with the filing dates of this patent. The patent had first been filed in 1942. Then the application was abandoned. It was refiled in 1947. If the initial filing had been kept on the books, then ENIAC's patent would have been dropped pretty quickly. But with the 1942 application dropped, things entered a kind of limbo. It's all a bit messy, but as near as I can tell, the Bell patent should have had priority over Eckert and Mouchley's filing. Bell would file their initial interference suit in 1959, and the court battle would take quite some time. And as with any legal proceedings, the actual interesting parts of the suit were dropped pretty quickly in favor of minutiae. By 1962, the case had been appealed, blocked, changed, retired, and eventually nearing

its end. The final case came down to a fight over a particular patent law, key being the idea of reduction to practice. Under U.S. patent law, you have a limited amount of time to file a patent on new invention. Specifically, you have a year after your invention is completed. Once you stop improvements and the device is in use, it's said to be, quote, reduced to practice. Sperry Rand vs. Bell Telephone ended up being all about when ENIAC was reduced to practice. The argument from Sperry, and the one that actually went out, was that ENIAC wasn't in use until the first firing tables were calculated on the machine. ENIAC wasn't in use until the first firing tables were calculated on the machine.

Firing table calculations were its original purpose, so it's reasonable to say that ENIAC wasn't done until it met its goal. That didn't happen until late 1946, meaning that the patent was filed within the allotted year. But Bell brought up a very interesting issue with that argument. When is a computer done? The judge's ruling on the case has a timeline of other alternate dates that were put forward by Bell. These include 1. Commencing in the latter part of November 1945, the ENIAC was used for computations on complex problems for the Los Alamos Atomic Energy Project. 2. After December 17, 1945,

the general principles of design and operation of the ENIAC were declassified, and only design details and circuits remained classified as confidential. 3. On February 1, 1946, ENIAC was used before members of the press to perform various computations, including a rerun of an operation of the Los Alamos problem. And the list goes on. ENIAC wasn't built to run calculations for the hydrogen bomb, but those were ran on the system before firing tables. Some details of the machine were made public before firing tables were ever filled? If it was declassified, dedicated, and then used for demos, then how is it still a work in progress? Ultimately, the case would only slow down ENIAC's patent process, but this shows the glaring hole in their legal armor.

The case concluded that both Bell and Sperry's computer patents were valid, but the issue of when ENIAC was completed will come up again. Just a few years after the Bell debacle in 1964, the ENIAC patent was finally approved. It took a full 17 years. But now, Sperry Rand held the rights to some of the core technology that made electronic computers possible. What would they do with all this power? As it turns out, nothing very good. At least, nothing good for the industry at large. Almost immediately, letters were sent out to all their competitors. Everyone who so much as touched a vacuum tube was now on notice. As of February 1964, anyone who made a computer was in violation of Sperry-Rand's patent. Either pay a licensing fee, or you'd wind up in court. It wouldn't be long before the patent was back in front of a judge.

In the intervening years between 1946 and 1964, the state of computing had really changed. Mainframes weren't just experimental devices tucked away in labs. It was an actual industry. IBM, Honeywell, GE, National Catch Register, RCA, Burroughs, they were all mass-producing computers. And so was Sperry Rand. There was an actual market with competition, something that didn't exist when the ENIAC patent was filed. Sperry Rand wasn't the only show in town. But with patent in hand, they now had a way to dominate any competition. At least, if everyone played along. Honeyball was one of the competitors that didn't want to play nice. Sperry Rand wanted royalties on every computer sold, something to the tune of 1.5% as a licensing fee. But Honeywell

didn't really buy it. They knew that the patent wasn't built on very solid ground. And frankly, no one wanted to suddenly pay royalties on their computers. Why would they? After what I can only assume to be a pretty angry correspondence, both parties decided to take their fight to court, and in 1967, suit was filed. Strangely enough, both Honeywell and Sperry Rand filed opposing lawsuits on the same day. The forms came in minutes apart, with Honeywell filing just barely first. So after less than two years, the ENIAC patent was back on trial. And this case would eventually invalidate the patent, and in doing so, some interesting information would come out to the public. The case would be known as Honeywell v. Sperry Rand, and it brought up three key issues with the ENIAC patent.

Well, one issue was with the patent itself, and two other were kind of administrivia things. So let's get the less fun ones out of the way first so that we can get into the more wild stuff, or at least what I think is a little bit more wild. One of the core arguments made by Honeywell's lawyers was that the ENIAC patent was filed over a year after the machine had been reduced to practice. It's the same argument that Bell made, but it was backed up with some new evidence. That new evidence came in the form of letters between researchers at Los Alamos and the Moore School. Crucially, these letters proved that ENIAC's first use was calculating

problems for the Manhattan Project, and that the entire machine was functioning from one wire to the other. From the court's decision, quote, the Los Alamos calculations which commenced December 10th, 1945 were the first problems placed on the ENIAC machine. When the first problem was put on the machine, it was the first time the machine as a whole was being used. It was fully expected that the problem would be solved, and it was. End quote. So ENIAC was completed and functioning prior to 1946. The court ruled that the task wasn't as important as the fact that it was up and running.

By December 10th, 1945, the design had been frozen. The machine had been built, and the first program had ran. On that day, the machine sitting at the Moore School was identical to the machine recounted in the patent filing. That meant that Eckert and Mouchley missed the deadline. On those grounds alone, the ENIAC patent wasn't valid. Legally speaking, that's enough. However, that's not the interesting part of the case. To me, the actual deadline of the filing is the least relevant. The second big problem came down, interestingly enough, to antitrust laws. And despite still being in the realm of legal mumbo-jumbo, I think this is a lot more fascinating. In 1946, Sperry Rand put most computer manufacturers on notice. But there were exceptions.

One glaring hole in that net was IBM. The reason? IBM and Sperry Rand were working together behind the scenes. Kinda. This goes back to earlier patent disputes between the two companies. During the 1950s, both companies levied multiple lawsuits against one another. There were a number of interference suits filed by IBM against the ENIAC patent, filings by Sperry Rand against IBM on grounds of patent violation, and a slate of mixes and matches of that formula. Both companies were very early players in the emerging computer market, so it really makes sense that they'd find some way to argue. market, so it really makes sense that they'd find some way to argue. But IBM wasn't just in court over patent disagreements. Big Blue was also dealing with the Department of Justice. It was

found that IBM was functioning as somewhat of a monopoly, partly in computing and partly in punch card tabulators. This went on trial in 1956, and IBM lost because they were actually trying to form a monopoly. There were a handful of legal repercussions, but the important one for us is in regards to their patent portfolio. Going forward, IBM was required to license patents to anyone who asked, and that included direct competitors. The sticking point here came down to patents and know-how regarding IBM computers. Sperry Rand actually lobbied the Justice Department to specifically include computing-related patents and information in the judgment. But that didn't work out. So while IBM had to license patents, they could still keep back some information.

So while IBM had to license patents, they could still keep back some information. Sperry Rand ostensibly wanted access to IBM trade secrets, because at the time, IBM was dominant in the market. Sperry Rand had Univac and a few other machines on sale, but they didn't have the same command and control of the market as IBM. But Big Blue's recent loss to the Department of Justice gave Sperry Rand a little bit of an opportunity. The two companies met outside of court and hashed out a deal. IBM was eager to clear themselves of any future patent suits, and Sperry Rand was key on getting access to some near-future IBM patents. They agreed to drop their lawsuits and draft up a patent-sharing agreement. Each company would get access to technical details from the other, and they'd also be allowed to freely use the other company's patents. A contract was drafted,

and the Justice Department even approved it. But come Honeywell's lawsuits, a few new details came out. The problem was that neither company ever intended to act on this new insider information. Eckert even testified in court that he had no clue what to do with IBM's quote-unquote know-how. Instead, the higher-ups wanted to use this agreement to form something of a legal barrier to entry for other companies. This is called a conspiracy. In 1956, IBM and Sperry Rand were the two biggest players in the game. By net sales, they made up 95% of the computer market. Their agreement let them hoard information between themselves. There's even evidence of internal talk and letting other manufacturers into the mix, but that was dropped in favor of maintaining a monopoly.

And with the agreement not to sue each other, IBM and Sperry Rand were freed up to pursue other legal prey. The final piece, the feather in the conspiracy cap, is that the duo worked to keep this kind of information under wraps. The Justice Department knew about it, or at least knew a lot about it. But crucially, none of the competitors knew what had been agreed to. Once details about the 1956 agreement were dug up in court, things became pretty clear. IBM and Sperry Rand were working together to stop competition, which is a violation of U.S. antitrust laws. Sperry Rand's selective enforcement of their

patent rights to ENIAC, well, that was just the tip of a much larger iceberg. What's strange with this part of the case is that the judge sided with Honeywell. They agreed that Sperry Rand was in violation of antitrust laws, but there weren't any damages paid out, and there weren't any direct repercussions for the company. If I'm reading the ruling correctly, then it seemed that the like treatment of antitrust violations was due to a statute of limitations issue. treatment of antitrust violations was due to a statute of limitations issue. That and the implication that Sperry ran didn't really benefit all that much from the agreement. It appears that over that decade, IBM won out. So that does it for all the legal arguments in the case, or at least

all the bickering over laws and legal statutes. That leaves just the good stuff. In other words, that leaves the fourth John in our story, one John Atanasoff, the legally mandated inventor of the electronic digital computer. Bell's complex calculator wasn't the only machine that came out to challenge ENIAC. To explain this, we need to take another jump back into the 1930s. During this period, Atanasoff was a mathematics professor at Iowa State College. But no matter the time and place, there's always one constant in life for a mathematician.

Problems crunching numbers. Atanasoff had access to desktop calculators and even an IBM tabulator. But despite being cutting edge, they didn't actually help that much. He was dealing with the same bottleneck that Mouchley would later describe. A simple calculator couldn't fully solve equations. Sure, it could make some of the number crunching a little faster, but you were still going to be doing most of the work by hand. Specifically, Atanasoff was trying to find a fast and easy way to solve systems of linear equations. Linear equations are commonly solved by converting them into a matrix, basically a big grid of

special numbers. Once you get everything fitting nicely in your matrix, it's just a matter of performing a few operations on the matrix itself. But that turns out to be a really repetitive and grinding process. At Nassoff figured that he could automate that away, but the tools of the day just weren't quite up to the task. He tried modifying one of Iowa State's tabulators, but the machine just wasn't flexible enough. Iowa State's tabulators, but the machine just wasn't flexible enough. He'd even tried building an analog computer to get the job done, but he was unhappy with its accuracy and speed. Over the course of trashing some tabulators, Atanasoff started to come up with an idea. He would need to make his own machine, something totally new. It had to be electric in order to maintain any semblance of

speed, and it would have to be digital in order to get any kind of accuracy out of it. His other big realization was that this machine had to use binary numbers, since that numbering system was really the only natural choice for electronic circuits. It was slow going, but in 1937, Atanasoff's vague notions would form into a concrete design. And after securing a grant from Iowa State, Atanasoff started construction on one of the first computers. Atanasoff and a grad student named Clifford Berry began construction in 1939. The machine was completed in 1942, dubbed the Atanasoff-Berry Computer, or just ABC. Once again, we aren't dealing with a computer in the truest sense of the word. ABC wasn't programmable, but it could reduce linear equations. In that sense, the

machine itself was one big specialized program. But the pieces were all there. ABC had memory. In this case, it stored data on a rotating drum. It had registers for storing numbers and results of calculations. And it performed calculations using logic-based circuits built from vacuum tubes. ABC would get close to working, but it never entered regular service. A series of two equations could be entered in via punch card and the computer could eliminate variables. Ultimately, it never worked to the scale that Atanasoff had planned for.

That being said, ABC was a huge step beyond existing methods. Atanasoff was starting to nail down how a computer would work. Despite being such an important machine, ABC wasn't that well known outside of Iowa State College. At least, the details of how it worked weren't really common knowledge. But in a strange twist, the onslaught of ENIAC-related litigations actually dredged up the earlier machine. Initially, during a 1950 trial, an IBM lawyer reached out to Atanasoff to testify. That would end up fizzling out when IBM and Sperry settled. But when Honeywell sued, a new series of lawyers got in touch with Atanasoff. And from there, a very interesting story started to emerge. You see, our man John Mouchley

was one of the few people outside Iowa State that knew about ABC. All the way back in 1940, before ENIAC was even a dream, Mouchley met at Nassau. First, at a AAAS meeting in 1940, Mouchley was giving a presentation about his work with early vacuum tube circuits. And, as fate would have it, at Nassoff was in the audience. After the talk, the two met. They struck up a conversation and would keep up a long-running correspondence about their respective research. The two were struggling with very similar problems, and they'd arrived at very similar solutions. It just happened that Atanasoff was further along. At the first meeting, Mauchly learned of the existence of ABC. The machine was still under construction at this point, but it was showing promise. Atanasoff was sparse on details, but he promised to show Mouchley more about the machine if he

would just visit Iowa State. In 1941, Mouchley made the trek to Iowa, and he came face-to-face with ABC. At Nassoff and Barry were both present to give a tour of the machine's current state. Barry even demonstrated it in operation. Mouchley would later claim that he learned nothing new from his trip to Iowa, even that ABC was underwhelming. But the fact remains that he saw the machine before ENIAC was designed. The two computers are vastly different. ENIAC is a strange, partly decimal machine while ABC is purely binary. ENIAC uses a combination of special purpose circuits while ABC is purely binary. ENIAC uses a combination of special purpose circuits, while ABC uses logic-based operations. But I can't personally believe that there was no influence from Atanasoff's work. In the sciences, you try to learn from the work of others,

even if it's just learning what to avoid. Mautley had to have taken something away from ABC, and that's just what Honeywell's lawyers contested. Their argument was that ENIAC was a derived work, that Mouchly used information gained from his trip to Iowa to make the computer. Outside of that, they argued that ABC existed as prior art. Even though Atanasoff didn't patent the machine, he was the rightful inventor of the electronic digital computer. They backed this up with expert testimony from Atanasoff himself and a series of letters sent between the two researchers. Crucially, it was shown that Atanasoff's work described the system very similar, at least on paper, to ENIAC. It used electrical pulses to ferry around numbers. It performed mathematical operations using vacuum tube circuits. And it did so quickly.

A small model of ABC was even built to explain its operation in court. Based off the evidence presented, the court ruled that ABC was the first electronic digital computer, and therefore ENIAC didn't hold that title. The court case would be grueling. A final decision wasn't reached until 1973. The final decision broke enough holes in ENIAC's patent that it was rendered invalid. This was a direct blow to Sperry Rand, but there were far bigger implications. Technically, this meant that Atanasoff had rights to the digital computer. But he had never patented ABC, so the invention fell into the public domain. In other words, no one really owned the idea of a computer.

This made the computer market safe from a lot of future issues. Sure, monopolies could still form, but by invalidating the ENIAC patent, it became a lot more difficult. There was no longer a single key to the computing kingdom. In the ensuing years, competition would drive the industry ever forward. Alright, thus concludes our series on ENIAC. Once again, we're left with an image of a complicated machine. The computer has a long-reaching legacy, both technically and legally. The eventual invalidation of its patent paved the way for future innovation, but it also helps to illustrate how complicated

computer history can be. ENIAC is the perfect example of what I mean when I say looking for concrete firsts is a bit of a wasted effort. Was ENIAC the first electronic digital computer? Well, there isn't really a satisfying yes or no. We can say that it was the first electronic digital computer that was programmable and used at any scale, but that's a whole lot of caveats. Legally speaking, ABC was the first computer, but it was never fully operational, and never even intended to be programmable. I think the shades of gray that we run into when looking at these sorts of questions are really a lot more interesting than trying to score points. But I also think we've zeroed in on at least one good answer in this series. Why did anyone want to make computers in the first place? We now have a pretty good sample size, at least I think so. Atanasoff hit on the idea in relative isolation.

So did Stibitz. The same is true of Mouchley and Eckert. The idea did come to Mouchley before he visited Atanasoff. They all had issues with mathematics, and they needed to be solved. They all wanted ways around running calculations. It was just too slow and too error-prone to use existing methods. Now, that's a little specific, but I think we can generalize. Throughout my research for this series, I kept coming back to something that Admiral Grace Hopper wrote. When explaining why she came up with the idea for a compiler, she would often say, quote, No one thought of that earlier because

they weren't as lazy as I was, end quote. Thanks for listening to Advent of Computing. I'll be back in two weeks time with another piece of the story of the computer. And hey, if you like the show, there are now a few ways you can support it. If you know someone else who would like the show, then why not take a minute to share it with them? You can also rate and review on Apple Podcasts. And if you want to be a super fan, then you can support the show through advent of computing merch or signing up as a patron on Patreon. Patrons get early access to episodes, polls through the direction of the show, and bonus content. Just last week, I put up our second bonus episode, this one on the history of OEM versions of DOS. So hey, go ahead, put in a dollar and check it out for yourself.

You can find links to everything on my website, adventofcomputing.com. If you have any comments or suggestions for a future episode, then go ahead and shoot me a tweet. I'm at adventofcomp on Twitter. And as always, have a great rest of your day.