Advent of Computing - Episode 65 - Teletype, Teleprint, and Telegrams

Episode Date: September 19, 2021

In today's episode we take a long hard look at the telegraph, and try to see how character encoding developed. We are dealing with 100% pre-computing technology, but there are some shocking similarities to later digital systems. Selected Sources: https://archive.org/details/electrictelegrap00highrich/page/2/mode/2up - Early history of the electric telegraph http://www.samhallas.co.uk/repository/telegraph/b6_baudot_multiplex.pdf - 1934 pamphlet on the Baudot telegraph https://ia800708.us.archive.org/view_archive.php?archive=/22/items/crossref-pre-1909-scholarly-works/10.1049%252Fjiee-1.1901.0058.zip&file=10.1049%252Fjiee-1.1905.0034.pdf - Murray's comprehensive article on telegraphy

Let's say you want to write the comprehensive history of computing, all in excruciating detail and, of course, all in chronological order. Where would you start it? This is a really serious question, and it's something that I find myself thinking about more often than I probably should. When I started planning this podcast, one of my first decisions was to not try to go in chronological order. A big reason for that decision was that I couldn't figure out when the starting point should be.

Let's say we restrict everything down to just examining electronic digital computers. Then the question turns into, what was the first machine to match that really narrow criteria? Legally speaking, there is an answer. The Atnasoff-Berry computer, built in 1942, was upheld in court as the first electronic digital computer. But just two years earlier, an engineer at Bell Labs built a machine that also used digital logic to do math. It was just missing things like branching conditionals. To cover that ground, which I think if we're talking about these electronic machines we have to, then we gotta push our start date back a little further. But what about non-digital computers? Analog machines existed well before 1940, and they were filling a very

similar role to later digital machines. The techniques and technology developed for analog systems were later adapted for more modern computers. So, shouldn't that also be included in the story? To cover that, you gotta push even further back. The real point of no return, at least for me, is when I start looking at fundamental questions. For instance, how did we figure out data encoding? Where did the whole idea of scrunching letters into binary numbers come from? Or hey, even something as simple as, why are keyboards the default computer interface, and why are they formatted the way they are? Those types of questions are especially dangerous because they push our horizon even further back. And I think that makes them

really important questions to answer. Welcome back to Advent of Computing. I'm your host, Sean Haas, and this is episode 65, Teletype, Teleprint, and Telegrams. This is going to be one of those back-to-basics episodes. We're traveling back to a time before computers, but we're still very much dealing with computer history. Now, let me just try to justify things a little bit here. Have you ever spent much time thinking about how characters are encoded on a computer?

Computers and machines built for running numbers, everything from memory elements down to processor pipelines are built to store and process numeric values. This even extends down to the hardware layer. Binary, the series of on-off pulses used by computers, was designed to represent numbers. This is a very mathematical ecosystem. So how do you cram letters into this mathematical world? What about punctuation? What about all the textual marks that you can't print, like a new line, for instance? If you're a programmer, then this quandary should be a very familiar one. The name of the game here is encoding. It's a series of schemes used to map letters onto numbers. But there are different

types of encoding for different types of text, or just for different types of computers. Some programs like ASCII, others use UTF-8, but at their core, it's all about mapping between numbers and letters. It's all about making an agreed-upon system to convert the written word to a series of on-off pulses. So where did this all start? Well, it comes down to the telegraph. First hitting the scene in the 18th century, yes, that long ago, telegraphs facilitated near-instantaneous communication for decades. The technology would mature from simple clicks and clacks all the way up to fully encoded text and automatic typewriters. Eventually, the same machinery was transformed into early computer terminals.

This borrowed technology would graft some quirks into computing that have stuck around ever since. If you've ever used the term baud rate or had to remember to print a carriage return after a line feed, then you can thank the telegraph. This episode, we're going to be diving into how these machines worked and, more importantly, seeing what later technology they would inspire. Along the way, I want to pay special attention to how familiar problems lead to strangely familiar solutions. It's an old story, but I think it's worth our attention. After all, what's more basic than turning text into pulses on a long wire? I guess the best place to start would be describing what exactly a telegraph is, since that's going to be the baseline the story builds up from. Specifically, we're concerned with the electric telegraph. The technology here is simple. You

have a long single wire that connects from point A to point B. Each end has some type of transceiver that can send and receive pulses on that line. And that's about it. The only smarts are really in the transceivers. The rest of the telegraph system is literally just a really long wire. The whole point of a telegraph is to transmit a message quickly over a long distance. That's something folk have been after for a long time, probably since language first emerged. In the 21st century, this type of communication is easy. We have satellites, we have networks, and we have cell towers. But for most of our history, fast, and I mean truly fast, communication was difficult.

For this opening section, I'm going to be drawing heavily from The Electric Telegraph, Its History and Progress, written by Edward Hyten in 1852. I did say we're going pretty far back, right? Hyten describes early means of long-distance communications as evolving out of a need for military intelligence. This starts with signal fires, flares, flags, and signal mirrors. To quote, The ancient Britons had their signal fires to warn the country of the approach of an enemy.

The Romans employed the same expedient in times of war, to telegraph their victories or defeats. End quote. Go ahead and squint at that a little, because there is something familiar hiding there. ahead and squint at that a little because there is something familiar hiding there. A signal for victory or defeat sounds an awful lot like a data encoding system with two possible values. It may be a little bit much to assume the Romans lighting signal fires were thinking in terms of binary data encoding, but I think this is an important insight. The whole idea of signaling a yes-no value is simple. It's the first thing people went for. You can't make flames send a very complicated message, but you can use them to transmit a simple yes or no, an on or off. With advances in the study of

electricity, everything would change. Or rather, a lot would change. One basic factor that I think we really take as granted when talking about computers is that digital signals just plain work well with electricity. Disregarding everything else, digital on-off pulses are a really natural media to send on a wire. If you only have some source of current and a wire, then the easiest thing you can do is generate on-off pulses. That's really all you can do. It ends up taking extra work and extra components to create analog variable voltages on a wire.

Digital is just plain simple. It's right there. But on its own, a battery and a wire aren't super useful. Maybe you can heat the battery up, but that's about all. Humans don't really have a way to naturally observe electricity. That's not one of our senses. Well, you can observe it by getting shocked, but that's just not a sustainable tool. Two early inventions that changed this were the electromagnet and the galvanometer. These are markedly similar devices. Both use the magnetic field induced in a coil to do something. In the case of a galvanometer,

that magnetic field is used to rotate a dial. This is often calibrated to measure electrical current on a wire. An electromagnet, at least generally speaking, is just a temporary magnetic field generated by a coil of wire. This can be used, for instance, to attract a needle when current passes through the coil. Both of these devices are converting an electrical impulse into some physical movement that humans can easily observe. Interestingly enough, they even predate practical light bulbs, so maybe instead of an idea manifesting as a light bulb over someone's head, it should show up as a pivoting needle. In practice, you don't even need a very large current or a high voltage to get a reading using a galvanometer or electromagnet. These two devices, plus a smattering of similar mechanisms,

opened up the possibility of an electric telegraph. The basic design for these types of systems went something like this. You start with a really long open circuit, just a wire running from some sender to a receiver. At the sender side, you have a power source and some way to close and open the circuit. At the receiving end, you have some way to show that the circuit's been closed, such as a galvanometer or some electromagnetic device. We can see the limitations of the system pretty easily. As described, you can send an on-off signal. That's cool and all, but it's not that much use on its own. Us language-using folk

have a much richer lexicon than just on and off. This basic template will remain consistent from here on out. The main area of research, at least the most relevant one on today's episode, is in how to formulate and tweak that system to facilitate sending more complex data. In other words, how inventors tried to send text using only on-off signals. This quest started as early as the 1770s. Once again, we're going way back here. It seems that the first area of attack was the transmission medium itself. According to Heighten, the first of these attempts was in 1774 by one Georges-Louis Le Saguier. And there's going to be a number of French names in here. I am very sorry, I'm trying to pronounce them as best I can.

Anyway, his device used 24 separate wires to transmit data. In this schema, each wire represented one letter of the alphabet. Well, almost the entire alphabet. Le Sagué didn't patent his device, and the only book I've seen cited that mentions his telegraph is in French and not translated or OCR'd, so I can't find super clear detail. It's either a misprint in Heighten, or some letters were dropped or doubled up. Some inventors in this period would follow a similar path as Le Saget.

These multi-wire telegraph systems did have their advantages. For one, you didn't really have to encode anything. You could just click the A button on the sending side, and the A indicator would click down on the receiving end. Now just do that for each letter of your message, and you're done. You've just transmitted data nearly instantly over miles and miles of distance. transmitted data nearly instantly over miles and miles of distance. So we're done, right? Just use 26 or so wires, give or take a few, and it works. Well, not so fast. It should be clear that using that many wires just isn't practical. For a demonstration, a bundle of 26 or 24 wires is easy to make, but imagine scaling that up. Even if you're just running a mile-long line, you would need at least 26 miles of wire. You would also need all the proper insulation to keep the wires from contacting each other,

and some kind of weatherproofing to keep corrosion or rust at bay. To get around the whole multi-wire issue, better encoding had to be developed. Well, I guess strictly speaking, any encoding had to be developed, since I don't think the 26-wire telegraph is really doing much encoding. Between 1774 and 1837, a number of inventors took the plunge to make a lower wire count telegraph. To simplify the story and prevent us from drowning in minutiae, I'm going to skip ahead and highlight two key examples. I just want to be totally clear here that there were earlier single-wire telegraphs and there were also many more multi-wire telegraphs than I'm going to cover.

As the first example, I'd like to present to you the Cook and Wheatstone telegraph. This device was built in 1837, hence the very specific end date to that range I mentioned earlier. It's not a single-wire device, but I think it's a fascinating example of how encoding was being tackled in this early period. The Cook-Wheatstone telegraph had five wires. The packing of these wires was pretty strange. At this point, insulated wires weren't the most common part at the hardware store, so a crude insulation was made using these wooden spacers. These spacers kept bare wires separated. Then everything was just crammed into a series of tubes and buried. This part doesn't impact encoding, I just think it's

kind of weird and neat. Both ends of the line were hooked into transceivers. As near as I can tell, there wasn't any hardware to negotiate who was sending and who was receiving. You just had to work that out on your own. Inputs here are simple. You get 10 buttons. These were wired up so that you could either send a positive or negative voltage down each wire. Hence, two buttons for each wire. The receiver, or more specifically the display, is what I find most fascinating here. It consisted of a giant engraved diamond shape, with five dials connected to galvanometers stretched across its midline. I'll link out to some pictures so you can see it more clearly, but these things look wild. The diamond was engraved with 20 letters placed on the intersections of a grid. To send a message, you depressed two buttons, closing the circuit on

two wires. The corresponding galvanometers on the receiving end then deflected. The two affected needles would then be pointing to a letter on the receiver's grid. This is all a little convoluted. To put it more simply, letters were encoded as polarized currents on two wires. As tempting as it may be to point to this as binary, we aren't exactly in that realm yet. Each wire technically had three states, off, plus, and minus. Cook and Wheatstone also didn't adhere to any real numbering system. The encoding here only really had context when used with these diamond and needle machines. That said, the two were able to reduce the number of wires needed to transmit a message. This was possible thanks to their improved encoding scheme.

But they made a trade-off. Like I said, this encoding only really made sense if you're using a Koch Wheatstone telegraph. It was built into how the machine functioned. An operator did need a little bit of training to get the system down, but in general, this was user-friendly. At least a little bit. You just had to read where the dials were pointing, and then figure out how to get the dials to point where you wanted them to.

The other big factor to keep in mind is that the Cook-Wheatstone telegraph wasn't a totally new device. I think Hyten gives the best explanation. Quote, it must be observed, however, that this telegraph contains little or nothing new beyond the particular combination of well-known parts. The use of the needle and coil was old. The employment for telegraphic purposes of galvanic electricity was old. The burying of insulated wires and tubes was old. The attractive force of soft iron to develop electromagnetic properties was old. End quote. This may initially just sound like a nitpick, but this is going to be crucial moving forward. Each new telegraph is an improvement on older devices.

Every new device is a remix or reshape of older technology. This is something that we should all be familiar with. Computers have followed the same course. In 1981, IBM didn't drop a totally new machine made from scratch. Their breakout hit, the IBM PC, was a combination of older technology, but put together in a new way. That means baggage. In this context, it means that no matter how far removed or futuristic a telegraph system may be, it will still borrow heavily from much older technology. be, it will still borrow heavily from much older technology. On the other end of the encoding spectrum, we have the Morse telegraph, and perhaps the better-known component of that system,

Morse code. This is probably the most successful single-wire telegraph system of the period. According to Morse, his big idea came to him in 1832, but the device wasn't built until 1837. This fits it snugly into the same time period as the Cook-Wheatstone telegraph. Morse's machine only used one wire. This really means one wire plus a ground, but for whatever reason, the convention just seems to be to not count a ground line. Transceivers were simple, almost deceptively so. You got a momentary contact switch to make and break connections on the circuit. Once connected, the circuit would

energize an electromagnet on the receiving side, which snapped down a spring-loaded arm. The receiver would just click every time the contact closed. Morse also built a recording mechanism that consisted of a metal stylus and a paper tape feed. As the electromagnet snapped down, the arm of the stylus left marks on the paper tape, thus turning electrical impulses on a circuit into a longer-lasting record. But, of course, the encoding is what makes Morse memorable. Morse initially called it a quote, system of signals that could be used to transmit letters over a wire. We know it today as, of course, Morse code. This code is

composed of so-called dots and dashes, dots being represented as a short connection on the telegraph's circuit, and dashes being longer periods of contact. The comparison to the Cook-Wheatstone system is pretty interesting here. Morse code takes time to learn. Everything has to be fully encoded, and you don't have some handy device that helps you encode or decode text. But there are some big benefits. For one, you only need one wire. Morse code can also exist outside the context of a Morse telegraph.

You can whistle in Morse code if you want. You can tap it out on a desk. You can flash it using a signal mirror or, eventually, transmit it over an audio channel. Since you only need one stream of data, it becomes versatile in ways that a multi-wire system just aren't. To break it down a little more, we could say that Morse code is a fully serial encoding. You get one packet of information followed by another and then another. All data is transmitted and received in the same order. This all sounds really basic, but I think it's crucial to understand. The encoding might be a bit of a hurdle to using Morse code, but the benefit here is that you gain a lot of flexibility in how you transmit and receive data. One last observation I want to make here is that compared to modern

methods of data encoding, Morse code has a bit of a problem. Despite having two states, dot and dash, a bit of a problem. Despite having two states, dot and dash, it's not really close to binary at all. We're missing the self-consistent numbering system aspect, but there's also the weird matter of time. In Morse code, an A is dot dash. A B is dash dot dot dot. Different letters take up a different amount of time. Data packet sizes are different. As we drive more and more towards automation, that makes for a bit of a little hiccup. The next big step in the evolution of the telegraph was to find a middle ground between encoding and ease of use. This is where we enter into the printed telegraph. The holy grail here was some type of machine that could be connected up using a single pair of wires, allow an operator to type in plain text to send, then on receipt could print out clear text.

Think of it as a middle ground between Morse's system of dots and dashes and the specialized machines being constructed by Cook and Wheatstone. Perhaps it comes as no surprise that this was actually really difficult to accomplish. The most common solution was to simply build machines that could automatically send out bursts of Morse code at the press of a button, and then pair those with a receiver that could decode Morse code into text. These ended up being complicated clockwork devices because, like I touched on earlier, Morse code uses a variable length encoding system. You need to be able to tell when one letter ends and another begins. Some machines interspersed signals between letters, but that started to diverge from

the accepted standard. In 1869, Émile Baudot entered into this strange transitionary world. That year, he started an apprenticeship at the French Post and Telegraph Administration. He was initially trained on the use of a Morse telegraph before moving on to something a little more challenging. Over a period of four months, he learned how to use and maintain a Hughes printing telegraph. This is what I mean when I say we're dealing with a learning curve here. The Hughes telegraph was one of those automatic machines that I mentioned. It used a complex mechanism to automatically generate Morse code at the press of a button. Incoming Morse code was then decoded and printed

onto paper tape as clear text. The whole variable packet length issue, for lack of a better term, complicated these machines considerably. Morse had designed his encoding system with humans in mind. More commonly used letters required fewer dots and dashes. E, for instance, is just a single dot, since that's the most commonly used letter in the English language. For the application of the time, this was a really good decision. It saves wear and tear on the operator's part, and it makes messages as short as possible, at least on average. However, there isn't a self-consistent logic to how characters are encoded. You just have to know Morse code,

and machines don't really handle that type of logic well. The Hughes machines that Baudot worked on were wildly complicated devices. I can't really make heads or tails of how they work under the hood. These contraptions really look like or tails of how they work under the hood. These contraptions really look like something out of an old science fiction novel. The Hughes Telegraph is built into a small desk. It uses repurposed piano keys for inputs. Outside the desk and the keyboard itself, most of the machine is made out of brass. All the gears and even its fancy governor are exposed atop the desk. As near as I can tell, it uses a type of encoding-decoding wheel to turn letters into dots and dashes and vice versa. The point is, we're looking at a fully mechanical and analog device. It's pretty easy for us to

pick this design apart. I mean, we have computers, just use one of those, right? Operators at the time also had their own set of grievances with the current art. Around the time Bado was getting up to speed at the telegraph administration, the main gripe was efficiency. Sure, it took an oddly variable amount of time to transmit a message. But more worrying, the telegraph lines sat idle most of the time. This wasn't necessarily an issue on the mechanical side of things, more with human use patterns. Operators weren't exactly tapping messages one after another for eight hours a day with no pause. There were gaps and there were breaks, even if it was just to flip a page. This idle time was a possible point of innovation. We don't know very many details about what sparked Badeau's interest in this area, but

we do know the rough strokes. A higher-up at the Telegraph Administration recommended that Badeau investigate ways to send multiple telegraphs on the same line. By 1870, a prototype was developed, line. By 1870, a prototype was developed, and in 1874, Baudot was granted a patent for his device. The machine used a new type of encoding and could send up to four telegraphs at once. There were three big keys to this technological marvel. Encoding, multiplexing, and at the heart of it all, a device called a distributor. These three factors are all really deeply linked together. It's a one begets the next kind of deal. Plus, just as an aside, since we don't have really good detail on the development process, we're going to just look at the finalized system. So let's start with encoding and then flesh the picture out from there. Bedeau's new telegraph used five bits to

represent a single character. Now, when I say bit, I really do mean bits here. Each character was turned into a series of five discrete on-off states. Specifically, at the time, Bedeau was using a positive and minus volt threshold for those states. These pulses were then transmitted and sequenced down the telegraph wire. Code tables showed each character as a series of five dots, empty for off positions and filled in for on. So while it's not strictly ones and zeros, this is a much better attempt at digital encoding. We're getting a lot closer to something we can recognize.

The rest of this new machine flowed from this one core decision. To us living in the fancy digital future, this choice seems trivial. Of course you'd want to go with digital as soon as possible. It's the best way to be. But let's examine this for a minute. What is Bedeot gaining here? For starters, each character in this encoding is the same length. It takes the same amount of time to transmit an E as it does a Z. There isn't any ambiguity there. Secondly, you only have one type of pulse to deal with. In Morse code, you need a dot and a dash. An automatic Morse code telegraph has to be able to generate and interpret both types of pulses. Bordeaux's code only needs a single pulse.

Call it a mono-dot system, if you will. The other little detail you need here is that Bordeaux wasn't designing his code to be transmitted by an unaided human. It was designed to be sent and received through machines. The name of the game here is standardization. By making everything as standardized and regimented as possible, Badeau was able to do some more advanced tricks. Perhaps the most impressive of those being multiplexing. This is how more than one telegram could be sent on a single wire at the same time. Well, let's just put some big scare quotes around at the same time. You can't send more than one

message down a single wire at the exact same instant. That's just how electronics work. That's how information theory works. Stuff gets jumbled. It's exactly like two people shouting over each other at the same time. The data is there, it's just not recoverable. What you have to do is jumble together those messages in some way that you can later reverse. Bedeau's solution was to interleave messages. If you're sending two messages, then start with the first letter of the first message, then switch to the first letter of the second message, then go back and send the next letter of the first, and so on. As just a quick example to hopefully drive the idea home, let's say I'm trying to send a message that's just AAAA and another that's just BBBB. The resulting data stream coming from Bedeau's new telegraph would be AB, AB, AB.

As near as I can tell, Bedeau was the originator of this idea, at least in practice. Nowadays, it's known as time division multiplexing. It's essentially a way to share a line by dividing up its time. As long as you know how the message is multiplexed, you can demultiplex it on the receiving end. Once again, standardization is key. This is only feasible thanks to fixed-length character encoding. The final piece of the puzzle, and what I think is the most amazing, is Bodeau's distributor. This is the heart of the telegraph. It's what makes everything work. The most succinct way I can describe the distributor is, it's a magic disk. And I'm

only half-joking here. The distributor is up there as far as physically simple devices that carry out complex tasks. For the most basic example, we're going to be sticking with duplex, as in two messages sent at once. Just keep in mind that everything can be scaled up to multiplex more messages. The multiplexer itself, the distributor, is just a carefully constructed disk that has a wire brush that rotates over it. In the case of duplex, the disk is broken up into two sections. In the case of Duplex, the disk is broken up into two sections. Each section is electrically isolated, so as the brushes rotate, the circuit alternately connects to one side than the other. Each section is further divided into five segments, one for each bit in Bedeau's code.

So as the brush starts rotating, it will read off each bit for this first channel. Bits are sent in parallel as operators only have to toggle in the bit pattern they want. Then as the brushes sweep by, it's turned into serial data. The first bit is read in, then as the brush moves forward, the second bit is read, and on down to the fifth bit. Once all five bits of the first channel are read, the brush enters the second part of the disk, and the process is repeated for the second data channel. Receiving just goes in reverse. The incoming signal is sent down the brush. As the signal pulses, the brush sweeps over the segmented disk. Timing on the receiving end has to line up with the transmitting end, of course, and in practice, all inputs and

outputs were buffered. The buffering at play here is also an interesting thing to consider. The operator input a bit pattern by pressing down five piano keys. Once depressed, they would lock in place until the distributor finished encoding their bits. Then the keys would spring back up, ready for use. This meant that an operator didn't need to be all that careful about timing. As long as they kept rhythm with the distributor's rotation, they could send out an uninterrupted stream of data. The receiving side used five electromagnets per channel to act as a buffer. Depending on the application, the received letters could be printed onto a strip of paper or

punched onto paper tape as raw code. For the time, this was a shining new achievement. Without laying new lines, a telegraph company could double, triple, or even quadruple their bandwidth. And best of all, it took advantage of idle time. That was a free resource, after all. But technologically, there's something just as exciting going on here. The distributor is doing two things. Well, maybe three if you want to be picky. First of all, it's running a parallel-to-serial conversion on the sending side and then a serial-to-parallel conversion on the receiving side. That's the trick of swapping between 5 bits represented all together and 5 sequential pulses of data on a wire. Second is the actual multiplexing.

By switching which channel gets attention, multiple streams of data can be mixed and then demixed. All of the operations only work because of the standardization in Bodo's code. Every character being the same number of bits and every bit being the same size, well, it turns out that that makes for much easier automation. But here's the part that really messes with me when I think about it. In practice, a group of operators would share one distributor. Quadruplex seems to have been the common choice here. You have four operators sitting at a table, each with their own keyset for entering bit patterns, all wired into one central distributor.

When you get down to it, each operator only gets control of the distributor one quarter of the time. So out of a minute, there are only 15 seconds where your keyset is actually connected and transmitting data. Does this remind you of anything? Perhaps something that comes up on the show a little bit too often? The way I see it, this is creepily similar to timesharing. Initially developed in the 1950s, timesharing is essentially a method where you switch between active tasks on a computer. Today, it's called multitasking. Back in the mainframe days, it allowed multiple users to share a single computer, but each user only had control of the system for a fraction of the computer's real time. It worked because of human use patterns. When you sit at a computer terminal, you aren't really hammering the machine

every microsecond. So there was some amount of idle time that could be exploited. Of course, there's a lot of differences here, primary being that timesharing was accomplished using software and pretty complicated software at that. That said, in the 1870s, Bodeau was implementing a similar idea using a magic segmented disk. I'm being a little fast and loose here, but we should be able to see at least similar shades. Timesharing and Bodeau's telegraph were both invented to solve similar problems. They both worked by taking advantage of underused resources. They both split up time between multiple users. To be 100% clear here, I'm not claiming that timesharing was developed by reinterpreting

old telegraph machines. I even went back to double check and I don't see any mention of telegraphs or multiplexing in the earliest papers talking about timesharing. What I'm more suggesting is that the ideas were out there. Not all problems and not all solutions are unique. Now, there are still some strange parts of the Bedeau system that need to be dealt with. This wasn't just some fully formed digital message passing system that was built in the 1870s. It had its quirks. A big one was the actual input method. This telegraph didn't use a keyboard in the sense of a typewriter keyboard. You didn't sit down to a nicely spaced out alphanumeric board. Instead, you got five keys.

These were used to input the five bits of raw Baudel code. The keys were divided into two groups, two keys for the left hand and three for the right. And when I say keys, I really do mean keys. These look exactly like keys off a piano. Operators were essentially playing chords on this tiny keyboard, but instead of music, they were creating electrical impulses. Like I touched on earlier, once depressed, the keys locked in place until the distributor finished transmitting. This meant that to be an effective operator, you had to keep up a rhythm so you didn't miss the transmission window. The result here is that operating Bedeau's telegraph took training and it took practice. Unlike the Cook

Wheatstone telegraph, operators had to memorize how to encode data. You didn't have some fancy visual aid. We're looking at something a little closer to Morse code in terms of learning curve. The automation aspect is really nice, but this comes with a trade-off. You can only use Bedeau's code and his multiplexing system with specific machines. You won't see someone trying to blink out some Bedeau code. That wouldn't work very well. Within this automated niche, Bedeau's invention really shined. The French Post and Telegraph Administration started using Bedeau's machines in 1877. From there, it spread throughout Europe and eventually overseas. I think one key to its popularity is that Bedeau didn't come up with this telegraph from scratch. Like earlier machines, it was a combination of existing ideas with some creative tweaks. The distributor was originally developed by Bernard Meyer and

used for multiplexing Morse code. At least, that's what I've seen repeated in a lot of articles. I can't find that much information on Meyer besides his connection to Bodeau's distributor system. And as best as I can guess, his distributor would have been substantially more complicated than Bodeau's later system, since Morse code is really its own kind of thing. The printing mechanism, which I haven't really talked about since we're really more on the data side of things, was also using borrowed ideas. It's remarkably similar to the paper tape printer used by Hughes' telegraph machines. I've also ran into a lot of sources that claim Bedeau's code had similarities to existing 5-bit codes. Specifically, that Bedeau borrowed from a

system devised by Johann Gauss and Wilhelm Weber. Once again, this may very well be the case, but I haven't been able to find the supposed code that was borrowed. Gauss and Weber devised some type of needle code in the 1830s. Think something like the Cook-Wheatstone system we talked about earlier. The examples of that code I've been able to find used variable length encoding, so maybe there's just something I'm missing, but I'm still a little dubious on that specific case. What is clear is that there were a lot of options floating around for telegraph encoding. Bado's code had unique characteristics, mainly its combination of serial data transfer and fixed character length. That said, we could almost look at the Cook-Wheatstone

system as 5-bit. It did use 5 wires after all. The point here is, the pieces of Bedeau's telegraph were floating around. So when it was unveiled, it must have seemed like a logical evolution and not so much a sudden leap. The last stop on our teletour is the teleprinter, or perhaps better put, the combination of teleprinter and teletypewriter that turn into the teletype. Ugh, a lot of tele-stuff today. As far as I'm concerned, this is where we start to see more than just shades of similarity with much newer technology. This is going to be the bridge from telegraph into more modern quirks of tech.

The key player here is going to be one Donald Murray, a journalist by trade. Now, I think this is probably as good a time as any to address the strange career paths everyone in this episode has taken. Barreau did in fact work in a telegraph office, but everyone else came into telegraphy as something of an outsider. Morse is a perfect example. He was a painter. Tinkering and inventing new devices seemed to have been a secondary outlet for many. I bring this up to point out that the telegraph was something of a casual technology once it got going. It was just part of people's lives, part of the world. So a painter was exposed to long-distance communications and became interested in making some tweaks. And in Murray's case, a journalist was exposed to long-distance communications and became interested in making some tweaks.

And in Murray's case, a journalist was exposed to the limitations of existing telegraphs and decided to make some tweaks of his own. Murray worked for the New Zealand Herald. He got that job in 1887. Then, just as now, communication made accurate and timely reporting possible. But 1887 wasn't exactly the best era for fast communication. Murray started out reporting on parliamentary sessions, but that required traveling. The New Zealand Herald's offices were located in Auckland, while New Zealand's Parliament is located in Wellington, New Zealand, that's some 400 miles away. Murray would type up his piece on the daily happenings in Parliament, then pass them off to a telegraph operator. The operator would convert

the clear text to Morse code and send it down the line to Auckland. Then a second operator on the receiving end would do the conversion back into clear text, usually typing it up on another typewriter. As near as I can tell, this was still all being done manually in New Zealand at the time, at least on the line Murray used. Of course, this was a really wasteful process. Just looking at the bill of materials alone, we have two typewriters, two trained operators, a reporter, and 400 miles of wire. Once again, we're seeing the issue of idleness and usability rear their ugly heads. As Murray saw it, this represented an opportunity. What if you could just replace the middle parts of the entire process? What if you could type a message on

a typewriter and have that transmitted to some remote printer. This is where the idea of teletype really starts to get brewing. Now, it really seems to me that Murray was the kind of person that just liked to be busy. That's something I can definitely relate to. During this period, he was working days for the New Zealand Herald, but he was also enrolled at Auckland University. After a handful of years reporting and studying in New Zealand, he switched jobs and moved to Sydney, Australia. There he started reporting for the Sydney Morning Herald and, simultaneously, started a master's program at the University of Sydney. After graduating with a Master's of Arts in Logic, he began his teletype investigation in earnest. He would start to see

results around 1892. Murray called his initial system the printing telegraph, but we've seen that term applied to a lot of different systems, so it's not the most specific indicator. Also, as we've seen before, it was a savvy combination of older technology and a few new modifications. it was a savvy combination of older technology and a few new modifications. One of the twists here is that instead of creating one monolithic machine to handle all aspects of telegraphy, Murray would build out a series of smaller devices. When used together, they allowed you to type a message into a keyboard and then send it to be printed on the other end of the telegraph line.

As you will see, breaking the task up into multiple smaller operations allowed Murray to save on some work. The big hallmark device here was, of course, the keyboard. Specifically, the so-called perforating keyboard. Murray's intention was to build a machine that anyone could use to type out a message, one that didn't require additional training. At the time, almost anyone who wrote knew how to use a standard mechanical typewriter, so he settled on copying that design. Those mechanical typewriters happened to use the QWERTY keyboard layout, at least most of them, so Murray's machine had its own QWERTY keyboard. What makes Murray's keyboard so different is that instead of throwing a hammer and printing a letter on a sheet of paper,

it printed perforations on a paper tape. Each press of the key punched a series of holes onto the tape, then advanced the spool to the next open location. So, typing on Murray's perforating keyboard turns your message into a series of perforations on paper tape. Pretty self-explanatory. What exactly were those perforations on paper tape. Pretty self-explanatory. What exactly were those perforations? Murray was using his own special code, which we can just call Murray Code, but it was really close to Badeau's earlier system. By 1899, the year when Murray filed patent for his system, this new code was fully evolved into a distinct system. It was a 5-bit fixed character length encoding,

just like Bidot. It was designed to be sent from machine to machine and worked with multiplexing. But the actual encoding, the mapping from bits to letters, was totally different. Operators were now 100% removed from the underlying encoding system.

This meant that Murray didn't have to worry about user fatigue or bit combinations that were too hard to press. Instead, the mappings were switched around so the most frequently transmitted characters used the least number of holes. The idea being that this would reduce wear and tear of the punching mechanism. For instance, E used only one hole, while Q used four. Murray saw the separation of operator and medium as a crucial factor. The best sourcing on Murray's work that I've been able to find is a 1905 article he wrote, simply called Setting Type by Telegraph.

In it, he explains, The Morse alphabet has been in possession of the field so long, and telegraph officials in that is not the case. All the operator has to do is learn typewriting. End quote. The benefits to obscuring encoding are twofold here. Most superficially, if you can hide away your encoding, it makes your system easier to adopt. For Murray, this meant tucking his 5-bit system underneath the keyboard. Besides just encoding, this also showed up in Murray's overall system operation.

Once you finished punching up a message to paper tape, you still had to actually transmit your data. This was handled by a separate transmission machine that read in your new paper tape. It turned that into serial streams of pulses, and it handled multiplexing and sending your message out down the line. This was another example of abstracting operators from raw data, but in a slightly different way. Think back to the cadence requirements of Bordeaux's telegraph. An operator had to hammer out characters in time to the distributor's sweep. A hiccup in that rhythm meant the outgoing message would have a blank, or wasted space. To get this down, it took practice. Murray's system was much more

synchronous. Once a message was transferred to tape, it could be sent off as a perfect stream of evenly spaced characters. Operator error didn't affect the final transmission nearly as much. If you ran a pretty tight shift with one person typing another feeding tape into the transmitter, then in theory you could be using 100% of your telegraph line's bandwidth. That means you could have no breaks and no hiccups in transmission. But the really cool part is that fully automating away encoding allows you to handle more complex data. This can be manifest in some unexpected ways.

Speaking more specifically, Murray started adding control characters to his encoding system. Specifically, Murray started adding control characters to his encoding system. These are non-printed characters that impact the formatting of your final clear text message. The prime examples are the line feed and carriage return, which, up to this point, hadn't really been part of telegraphy. These new formatting characters worked in conjunction with the final big piece of Murray's system, the printer. This was essentially just a typewriter's printing mechanism hooked into a paper tape. The tape was then fed into the printer to be decoded to clear text. When the printer hit the end of the line or received a line feed slash carriage return, the receiver would reposition and start typing at the beginning

of the next line. So line feed and carriage return entered into the picture for a very specific use case. They started to make sense if you had a telegraph that sent synchronous data that was intended to be printed on a full-size sheet of paper. The last distinction matters because, as we've seen, a lot of other automatic telegraphs would print letters onto paper tape. It was just a simpler mechanism to implement. The wild part here is that we still use these two formatting characters today. The file I'm writing for the script has carriage returns all throughout it.

Every document you've ever written on a computer, every terminal session you've ever dealt with, has had carriage returns mixed in. Depending on the software, maybe there were line feeds too. Here's another weird one. If you're at a desk, I want you to look down at your keyboard, or if you're on a phone, then pull up a virtual keyboard. In most parts of the world, you'll probably see an alpha block that starts with Q-W-E-R-T-Y, or QWERTY. That's the same layout that Murray chose for his telegraph. This is probably the most distinctly visible legacy we can see from this era of the telegraph. The reason for the spread, or at least the jump to computing, is pretty simple. Murray's telegraph system became wildly popular. In 1901,

Scientific America ran an article on the quick and easy telegraph, and that was really just the beginning. Within a few years, Western Union adopted the system, so Murray's machines started seeing widespread adoption. The encoding Murray used also became a firm standard in the world of telegraphy. In 1924, it was codified as the International Telegraph Code 2, also known as Bodeau-Murray Code. Along with the success came improvements, both for Murray and others. We start seeing combined systems as early as the 19-teens. These new devices, often just called teletypes, basically just smushed Murray's series of devices into one large machine. You had a keyboard, tape puncher, reader, transmitter, receiver, and printer all in

one package. Crucially, some of these machines allowed you to bypass the intermediate step of producing punched paper tape. Instead, the keyboard translated keypresses directly into serial data. Incoming data was also printed directly onto a paper feed instead of first being committed to paper tape. What we start to see for all purposes are digital text transceivers, machines that can be used to enter text in a digital format and then display text translated from that same digital data stream. Of course, this all developed prior to digital computers, meaning this I.O. technology was in place as computers were first being developed. These machines were a tried-and-true technology, something that almost faded into the background of an office or a lab. And they already had well-established standards.

The first crossover we see happens in the very earliest days of digital machines. In 1940, George Stibitz, an engineer at Bell Labs, started investigating the use of logic circuits for carrying out mathematical calculations. His first big success was a device he called the complex number calculator. It wasn't quite a computer. Instead, it was a machine that used logic circuits to run calculations. The human interface that Stibitz went with, the proverbial face of his new machine, was a teletype. Digital machines only really work with other digital machines. Everyone has to speak the same language.

And it turned out the teletypes were ready and waiting. With the press of a few keys, equations were sent into the complex number calculator, and they were all encoded as Bedeau-Murray text. Relays fired, some lights flashed, and outputs came back down the line to be printed on a paper feed. Everything was complete with a QWERTY keyboard and formatting characters. Alright, that does it for this look into some pre-computing technology. Going from the early days with Telegraph up to simple encoding, more complex encoding, and eventually early terminals. Here, as we close out, I want to give one final justification as to why looking at this type of technology is important for understanding computers.

The history we've been covering in this episode has been tracing progression that long-time listeners should be somewhat familiar with. When you get down to it, this story is all about abstraction. Not just the idea of getting people further and further away from underlying data, but also figuring out a reasonable way to accomplish that. Some ideas don't really go anywhere, like a 24-line telegraph, for instance, but in route, a lot of rough edges are worked out. Buffering, something that's shown up in most of the systems

we covered today, is a surprisingly important example of this. We saw this in Bedeau's Telegraph, how single characters would be held in buffer as the distributor made its passes. Murray took this to the next level by committing entire messages to tape before they were transmitted. This is a technique that's crucial to modern computer systems, especially communications technology. In the examples we saw today, it helped simplify and synchronize telegraphs. In the more modern era, buffering can serve a similar purpose. Audio streams will be buffered to make playback more smooth. Keystrokes are often buffered to simplify hardware I.O. It's a common practice today,

and it was just as important in the era of the telegraph. Data buffering is just one example. What I'm getting at is you should keep your eyes open for more. You never know when you're going to run into some thoroughly obsolete technology that's hiding some familiar tricks. Thanks for listening to Advent of Computing. I'll be back in two weeks' time with another piece of computing's past. And hey, if you like the show, there are now a few ways you can support it. If you know someone else who'd be interested in the story of computing, then why not take a minute to share the show with them?

You can also rate and review on Apple Podcasts. And if you want to be a super fan, you can now support the show directly through Advent of Computing merch or signing up as a patron on Patreon. Patrons get early access to episodes, polls for the direction of the show, and bonus content. You can find links to everything on my website, adventofcomputing.com. If you have any comments or suggestions for a future episode, go ahead and shoot me a tweet. I'm at Advent of Comp on Twitter. And as always, have a great rest of your day.