Advent of Computing - Episode 69 - The oN-Line System, Part 1

Episode Date: November 15, 2021

NLS, or the oN-Line System, is often looked at as a mile marker in the development of modern computing. It was the first system to use a mouse, one of the first functional examples of hypertext, pioneered remote collaboration, and so much more. But how much do you know about NLS itself? In this series of episode I'm picking apart the system behind the legend. Part 1 deals primarily with the early roots of NLS, Augmenting Human Intellect, and Doug Engelbart's vision of hypertext. Surprisingly, a lot of this episode has to do with punch cards and a more obscure related technology: the edge notched card. Selected Sources: https://dougengelbart.org/content/view/138 - Augmenting Human Intellect https://americanhistory.si.edu/comphist/englebar.htm - Engelbart Oral History, with the Smithsonian

Have you ever felt a little overwhelmed? That modern life is simply too full of information? Maybe even that there's too much data in the world for you to process? Or that there's just too much data for any one person to know it all? Has this overwhelming feeling, perhaps, led you to underperform? Has it lessened your productivity? If you answered yes to any of these questions, then you may be interested in human augmentation technology.

Now, this isn't a very new feeling. Throughout our history, humans have always been frustrated with our own limitations. But there's a remarkable thing about us humans. We're supremely gifted when it comes to finding ways around our limits. Have too much stuff to carry? Well, let me introduce you to a little thing called the wheel. Getting rained on too much? Have you heard the good news about homes and shelters?

Perhaps one of the greatest limitations we face is also the hardest to work around. That being the human mind itself. Just prior to the dawn of the digital age, this limitation started to be a topic of some concern. Simply put, the specific issue was that there was too much information in existence for an unaided human to navigate. We can only read so fast, and we can't instantly scour an entire library. Of course, humans are good at finding ways around our limitations. The first workable solution came about thanks to the digital revolution. It was called the online system, or NLS.

Welcome back to Advent of Computing. I'm your host, Sean Haas, and this is episode 69, The Online System, part one. Today, we're finally getting back to my ongoing series on hypertext and the idea of digital utopia. In this episode, we're looking at the revolutionary online system and seeing how it links up with the larger story of hypermedia. But before that, I need to get some rationalization out of the way, at least to ease my own mind. I've had this running debate with myself as to if I should do this episode or not, and if so, how I should approach it. Way back when I started Admin of Computing, I had three topics in mind for my

first episodes. The Intel 8086, punch cards, and Doug Engelbart's mother of all demos. Of course, the mother of all demos was where NLS was first shown to the public, so there's some overlap going on here. Those ended up being episode 0, which never aired, and episodes 1 and 2. The 8086 episode eventually got unfurled into a long series on Intel's processors, the Punch Card episode became the first audio content that I actually published, and the demo was the second. My early episodes are… well, at least I think they're a little bit rough. I was still getting my podcasting legs, as it were.

Episodes were shorter back then, they didn't have the same level of detail, and they didn't really have the same production quality that I'm comfortable with. I like to think that I've learned at least a little bit since then. I've worked hard to stick to a rule of not trying to remaster old episodes, since I think that would just lead to an unending cycle, which wouldn't be healthy for production or myself. But I've agonized a little over that rule as pertains to episode 2 specifically. Since I started my current sporadic series on hypertext, I knew I'd eventually want to come back and talk about Doug Engelbart in detail.

The other reason that I've known this would eventually happen is that I've been doing some really extensive research in the background. You see, a while back I made a pilgrimage to dig through a certain special collection at a certain university, but we're gonna get to that. This episode isn't going to really be a repeat of episode 2, at least I'm trying my best to avoid that. Instead, we're going to be looking at Engelbart's vision of hypermedia, where his inspiration came from, NLS's implementation specifically, and how this was all developed. Now, I want to place NLS in the larger context of the systems it inspired. I want to place NLS in the larger context of the systems it inspired. In 1968, the mother of all demos really lit a fire in a lot of minds.

The technology unveiled in that presentation would have radical impacts on computing all the way up to the modern day. I mean, that's the first time people saw a mouse being used on a computer. It's a big deal. And I think it's fascinating to look at how Ted Nelson's ideas on hypermedia, for instance, are in dialogue with Engelbart's earlier system. Or how later hypertech systems deviated from the trail blazed by NLS. But here's the thing. The online system is a really complicated topic to approach. I don't just mean technically, there's also a lot of theory and context behind the scenes that we're going to need to really make this endeavor worth

the trouble. So to that end, we're facing down a two-part series here. Let's call this NLS November because this whole month is going to be all about NLS. To kick things off, in this episode, we're going to be taking a deep dive into the background that leads us up to NLS. That means, strangely enough, a pretty long discussion of different types of paper-based storage medium. Now, I swear, I'm not joking here. If you just stick with me, this will all make sense. Next episode, we'll get into the real meat of NLS's

implementation and where that technology would go. Well, I guess this is where I should say, buckle up. For this episode especially, I'm going to be digging into some of my secret mind-of-madness research stash and taking us back to some of the basics of data management. Because, hey, what's more fundamental to hypertext than organizing information? Like most discussions of hypertext, we have to start with Vannevar Bush's As We May Think. I know, if you're a long-time listener, this probably sounds like a broken record. The whole point of what became known as hypermedia is to improve the human condition through better data

management practices. In the summer of 1945, Bush laid this general idea out in writing. As we may think, makes the argument that we've hit a point where the sum total of human knowledge is just way too much for a person to handle alone. He called this the information problem. The solution he offered was called the MIMEX, a theoretical machine that organized data into chains of linked ideas. That same year, 1945, the Second World War was drawing to a close. A young Doug Engelbart was sitting in a shack in the Philippines, manning a radar station and waiting to be shipped back stateside. During this downtime, he often took to reading, and one day he got his hands on a copy of The Atlantic. This particular copy would have probably been a few months old by the time Doug got his hands on it, but hey, it was new to him.

Leafing through, he came upon As We May Think. Years later, during an interview at Stanford in the 1980s, Engelbart still looked back on this moment as transformative. Quote, I remember being thrilled. Just the whole concept of helping people work and think that way just excited me. I can remember telling people about it. I never forgot that. End quote.

The information problem itself wasn't new. It's the kind of issue that humans had been facing since we started writing stuff down. But Bush's articulation of the information problem and his solution was something totally unique. What's more, as we may think, wasn't some dry academic work. It was published in a normal magazine. It was intended for a general audience. Bush was a really savvy operator here. He wrote, as we may think, to be accessible to just about anyone. The article reads more like a work of near-future science fiction than anything. It was intended to light a fire in the minds of the curious, and that's exactly what

happened here with Engelbart. Before we continue, we have to make a digression into some of the more technical content that Bush left out of his popular work. As I keep reiterating, and I will keep reiterating, the information problem wasn't some new crisis that appeared in 1945. Bush's Mimex wasn't the first pass at a solution to these similar types of issues. It wasn't even Bush's first attempt. At its core, the Mimex used a device called a rapid selector. This was an actual machine that Bush had developed prior to World War II. It allowed a user to quickly search through a reel of microfilm and pick out slides that matched some selection criteria, whatever that may be. The Mimix, which was only a theoretical device, paired this searchable storage with a slick interface and

a few more bells and whistles, but the crucial technology at work here was the rapid selector. crucial technology at work here was the rapid selector. So how did the rapid selector actually work? Well, all things considered, I don't think it worked very well. Each slide on your microfilm reel was split into two parts. One region of the slide had a duplication of an actual document, while the other section had a pattern of dots. These dots encoded basic information about the corresponding document. Things like a section of the title, the author, and maybe some categorical data. To search your microfilm, you first encoded your query as a pattern of dots that would match this schema. Then the rapid selector would scan through the reel of film looking for slides that matched your query.

The immediate technical problem here should be clear. This is all sequentially accessed storage. At the time, this being somewhere in the late 1930s or so, microfilm was one of the best ways to store large volumes of duplicated information. An entire shelf of books could fit on a single reel of film. In that capacity, it worked remarkably well. Each slide on a reel was a perfect photographic duplication of a single page. You even got the exact same structure as a book. One page stored after another after another. But that's all you got. The only upside compared to paper books came down to size.

That's where the rapid selector came into play. Bush was trying to solve the issue of finding specific information on a reel of film, but at the same time was restricted to the limitations of the format. That's something you always have to deal with. Format and media are going to be a fundamental restriction to anything you build. So that's one problem. The next big issue here was the data encoding that Bush used. If you look closely at a slide from a rapid selector, you'll notice that the pattern of dots looks more like a pattern of little rectangles. notice that the pattern of dots looks more like a pattern of little rectangles. You see, this data encoding made a pretty creative use of punch cards.

To encode searchable data for the rapid selector, an operator would punch up a card with the proper information. To query the rapid selector, you would just punch your query into a card, feed that card into the machine, and wait. You would just punch your query into a card, feed that card into the machine, and wait. The rapid selector made actual selections by comparing the query punch card to each microfilm slide's miniaturized punch card pattern. Once again, at the time, punch cards were one of the best ways to store and encode data. Specifically, it was really the only way to store digital data. In the 1930s, we aren't really dealing with the digital revolution quite yet, but there were still

a lot of reasons you'd need to have some type of discrete data storage. Accounting was one. At the time, punch card tabulators were used pretty aggressively for things like, you know, tabulators were used pretty aggressively for things like, you know, payroll, census data, simple data reduction, and a slew of other behind-the-scenes office-y and research-y stuff. The problem here was a little more subtle than the issues of microfilm. A single punch card doesn't hold a lot of data. A standard IBM 80-column model can encode, well, 80 characters of data, as the name would suggest. If you were running Twitter with early IBM equipment, you'd have to use four punch cards to compose a full tweet. That's not really wonderful storage density, but each card was just paper, so you could buy as many as you wanted.

In other words, any usefully large dataset required a correspondingly large number of punch cards. The implication for the rapid selector was that each slide could only have a small amount of metadata describing it. That on its own is a pretty bad limiting factor, but this can get a little bit weirder. So if you'll humor me, let's get a little weird with punch cards. Cards at this point aren't being used with computers. There just aren't digital computers to use them.

That said, you can't use a punch card on its own. You still need some type of equipment to punch and read them. No one is going to really sit down with a stack of a thousand punch cards and start sorting and tabulating them by hand. You need a machine to handle this kind of data. In a vacuum, that's fine. You have to use a tabulator, a card punch, a sorter, and some other equipment to really get the best use out of these

cards. But nothing occurs in a vacuum, especially not the boring behind-the-scenes stuff. While each punch card is cheap, the actual tooling to use them is not. Buying into a punch card automated office would be roughly equivalent to digitizing an office a few decades later. This is probably a hand-wavy comparison, the numbers probably don't 100% line up, but we're looking at a similar scale of investment and change. So a machine like the Rabbit Selector has to exist in this larger office ecosystem. I know it's an overly dramatic way to put it, but we could almost call punch cards a tool of the bourgeois. The last weird step on our punched path is this. Despite each card standing in isolation, punch cards are still a sequentially accessed medium.

I think this part is a little bit tricky and probably sounds a little boring, but stick with me here. Each card represents some single unit of data, a kernel of information. It could be an invoice, or the metadata for a slide of microfilm, or it could be an employee record. So, at least conceptually, that sounds pretty random access, right? Let's say I need to check on my power bill from last month. I just need to pull out a single card from the stack. But in practice, you can't do that. Looking at a stack of punch cards gives you no information about what's contained on those cards.

It's just a pile of paper. In a stack, you can see the top card and you can see the edges of every card after it. It's the same as looking at a book. You see the top page and then just uniform edges for every other page. Books have things like indexes and page numbers to help you find specific information, but that doesn't work with punch cards. These cards aren't bound, so they can have any order. So how do you find last month's power bill in this kind of mess? The answer is simple, as long as you remember this is a medium of the bourgeois. I swear I'm not just being grandiose here, there's a point to this analogy. You have to use a machine.

Specifically, you'd probably use a punch card sorter to order by the month field and then pick out the proper card. The machine will have to scan through each and every card looking for the right bill. In other words, we're firmly in the sequential access domain. The links that Bush describes in As We May Think do shift things around somewhat, but we're still in the sequential world. A link would reference some slide by an ID number. Then the microfilm would have to be scanned for that specific ID. All this has to be done sequentially. Despite being one of the first descriptions of a hypertext-like system,

the MIMEX doesn't actually support fully random access and fully free association. It's firmly rooted in a sequential way of thinking. This isn't a dig at Bush, more the limitations he was working with. Random access for large amounts of data doesn't really become a thing until we're well into the digital age. If you read As We May Think on its own, you don't run into any of these details. This is from Bush's larger body of work. All you get is the idea of a trail of thoughts. It's unbounded by the actual technical issues. That's why the essay is so powerful, while the mimics itself

never escaped its pages. I think this may have actually been a good thing, since this let Bush's ideas speak for themselves and weave their own kind of magic. Doug Engelbart was very deeply affected by this particular spell, but its full effects took some time to become apparent. by this particular spell, but its full effects took some time to become apparent. A few months after coming into contact with As We May Think, Engelbart was back stateside. He completed a degree in electrical engineering, then he settled into the all-too-familiar cycle of work and home life. That lasted until the end of 1950, and then, in Doug's words, he had a bit of an overreaction. He had been working a pretty mundane engineering gig, he only had a bachelor's degree at this point, and in his current situation, he could just keep coasting on as an engineer as long as he wanted. But he wanted

to do something bigger, something more important. Engelbart wanted to tackle some big problem. And by the beginning of 1950, a plan started to form. Quote, Somehow I said the world's getting so complex and everything happening was urgent and our ability to cope with that is not increasing as fast as complexity and urgency is. And that can only spell a higher and higher probability of global disaster if we don't do something. He continues, Then I got the image of the way computers could help interactively, and the radar technician training let me realize easily that a computer could make anything happen on a display screen. In other words, Engelbart decided to set his sights on the

information problem. But this wouldn't just be a straight implementation of Bush's vision. By the 1950s, there was a new hot technology on the block, the computer. And this is where we start to see Engelbart's visionary streak shine through. You see, at this point, Doug had never seen a computer. He had never used a computer. He was working in California, and the nearest computers were thousands of miles away from him. From limited information that he had access to, Engelbart somehow realized that these new digital machines could make the mimics a reality. In order for Engelbart to enact his grand plan, he had to continue his education. In 1951, he enrolled in UC Berkeley's grad program. He decided to stick with electrical engineering,

earning a PhD by 1955 and briefly teaching as an assistant professor at Berkeley. As near as I can tell, it was also around this time period that he was introduced to a crucial influence. For this next part to make any sense, we're going to need to fast forward a number of years. In 1962, Engelbart published Augmenting Human Intellect. As far as I'm concerned, it's his theoretical magnum opus. It laid out his thoughts, research, and plans for solving the information problem. Unlike As We May Think, this was a lot closer to a scientific paper. But of course, there's some caveats there. It's long, it's detailed, it's backed up by a little more experimental work than As We May Think, and it's really everything you's backed up by a little more experimental work than as we may think,

and it's really everything you could want as a starting point. There's a lot to be said about the paper. Engelbart was, before it was ever really possible, showing how computers could be adapted for a much more modern use. He was past the huge number crunching machines and into a more complicated future, but we'll get to that part later. Importantly for me, Engelbart goes into detail about his rationale for certain decisions and ideas. Tucked away in one of the paper's latter sections is a small passage that, when I first read it, sent me reeling. This is from a section simply called

Some Possibilities with Cards and Relatively Simple Equipment. Quote, A number of useful new structuring processes can be made available to an individual through development and use of relatively simple equipment that is mostly electromechanical in nature and relatively cheap. We can begin developing examples of this by describing the hand-operated edge-notched card system that I developed and used over the past eight years." Now, Engelbart spins the rest of this section explaining how he used edge-notched cards to store notes and formulate ideas while working on his research. It also goes into detail on how the specifics of the medium informed some of his thinking

on data association and linking. This would be my entry point into what's become a really deep rabbit hole. So why is this so important? Why did this seemingly mundane passage grab me? And what exactly are these cards that Doug's talking about? Edge-notched cards have also come up on the show a little bit before. Back when I was doing public bonus episodes, I did a short blurb on this technology. The easiest way to explain things is that they're a close cousin to punch cards, but that kind of sells the technology short.

You see, edge-notched cards aren't used for storing digital data. They're used for storing metadata and associations between points of other data. Here's how they work. Each card itself is a simple sheet of stiff paper. It's usually made from the same type of stock as punch cards. The middle part of the card, its face, is either left blank or printed up like a form with little regions for you to add data yourself. This is where the actual information goes.

It's usually in the form of simple text. Around the perimeter of the card is a series of round punched holes. Data is encoded on the card by cutting holes into notches, hence the term edge-notched. It's this pattern of notches that's really the key to this technology. Now, edge-notched cards started to develop just after the first punch cards. We're talking about the 1890s and very early 1900s here. The first patents on the technology were a simple adaptation of punch cards. Even at this early juncture, inventors saw an issue with punch cards.

It was the same issue I explained earlier. How do you pull a specific card out of a stack of punch cards? The general idea started out as adding notches to the edge of cards so, at a glance, you could pluck out certain cards. At first, this was designed mainly for categorization. The first patent on the technology issued in 1896 offered a way to visually categorize census data. One of the examples used in the patent was a set of notches corresponding to employment status. That way, a census worker could quickly grab cards just for farmers, for instance. Over the next few years, the technology was adapted for record-keeping. In the early 1900s, there are a number of patents that include notching

date codes onto the edge of index cards. This way, I could, in fact, pull out last month's bills with relative ease. Where this really gets going is in the 1920s. By this time, punch cards were well established as a data storage medium. Tabulators and all the other machinery needed for their use was in place. This is also when edge-notched cards become a fully divergent technology. A series of patents around this era show cards with the full pattern of holes and notches that I described. And they add another important detail. A knitting needle. described. And they add another important detail, a knitting needle. Now, this may sound funny, but check this out. This is what makes edge notch cards so different from punch cards. You don't use machinery to handle notch cards. They're a fully manual technology. Instead, you use a long

metal needle to make selections, and it really does just look like a knitting needle. Here's how this works in practice. Let's say I have a stack of Notch cards that are all of my power bills. Hundreds of them. I use a lot of power. I want to just get the bills for every December, since that's when I run all my space heaters, so I need to add up all the big numbers. In this filing scheme, I would have set aside a chunk of 12 holes to represent months.

So I'd find that region, find the 12th hole, and then thread my needle through it. Then all I have to do is lift the needle up while loosely holding the deck of cards. Any cards that I've notched out the December hole for will fall off the needle into my hand, and every other card with an intact hole will stay on the needle as I pull it up. That one operation and my selection is complete. Unlike punch cards, edge-notched cards are totally random access. You can pluck out any card or group of cards that you want, access. You can pluck out any card or group of cards that you want, provided you know the right notches. And unlike punch cards, this isn't done using a machine. In that sense, we could say that edge-notch cards are a tool of the proletariat. They only cost as much as a stack of paper and

a needle. They require no machinery, and thanks to their analog nature, you don't even need to make them with really high tolerances. You can very easily make a pack of notch cards using nothing but a pair of scissors and a hole punch. Edge notch cards hold this awkward place in the history of data management. There's no other way to put it. They are not high-tech at all. It's some of the most low-tech stuff you can deal with. They weren't spearheaded by some huge company like IBM. Edge-notched cards never made the leap into the digital age like punch cards did. So not only are we looking at boring back-office stuff, but we're looking at really cheap and simple office stuff. There's some interesting implications from this, and we'll address some of those as we

continue down this spiraling path. The big one that I just want to throw out right now is that edge-notch cards aren't very well documented. They're almost this kind of ephemera. Despite that, these strange little cards show up in many unexpected places and survive well into the computing age. Putting that aside, well, aside, I want to focus on the encoding. I was using a simple month-notch as an example, but it doesn't have to even be a single notch. The encoding system can be as complicated as you want to make it. So let's meet back up with Doug Engelbart. From that appendix in Augmenting Human Intellect, we know that he was using edge notch cards starting somewhere around 1954 or so. This is an early formative period, when he knows that he wants

to tackle the information problem, but is still working on how to accomplish that big goal. So how do edge-notched cards fit into this larger picture? Well, you see, I may be the only person in a position to actually answer that question. Like I said earlier, I've been fascinated by edge-notched cards for a while now. You can ask any of my friends, and they'll tell you it's kind of become an obsession for me. So when I learned about Engelbart's connection, I just couldn't leave well enough alone. I did what I usually do to answer these kinds of questions. I started harassing some archivists. After Engelbart's death, his family donated most of his notes and files to Stanford's special collections. That's a pretty common occurrence for a lot of high-profile

figures. It's a way to preserve raw information for later use, usually by people like me. The problem is that these collections aren't always very well organized or documented. Now, that's no knock on the archivists that handle special collections. Far from it. They just have too much stuff to deal with. We're talking warehouses and warehouses full of papers, books, manuscripts, notes, and, hopefully, edge-notched cards. I was able to get in touch with Christina Engelbart, Doug's daughter, and verify that if the cards still existed, they'd be somewhere in that collection. From there, it was just a matter of finding which boxes in the collection looked suspicious, filling some requests, and then making the

road trip out to Stanford. It took me a couple of days of rifling through boxes and folders and a few paper cuts, but I was able to find the cards that Engelbart mentioned in Augmenting Human Intellect. At least, I was able to find a lot of them. To be honest, it was an exhilarating experience. I definitely hit the city to celebrate that night. What I dug up was fascinating. Not only was Engelbart a really hardcore user of edge-notched cards, but he used them in some unexpected ways. And if my uncounted hours of studying these cards is correct, I think the medium impacted Engelbart's later work, which means that, speaking on just the lineage in general, these cards are a missing link to our understanding

of hypertext. To start with, we have the matter of encoding that I keep bringing up. Most decks of cards that Engelbart kept around were for notes, contacts, or just scribbles of information. Interestingly enough, some even had newspaper clippings that are literally cut and pasted onto cards. Now, accordingly, Engelbart called these note decks, since, you know, they were decks that had notes in them. These offer another example of the most simple kind of encoding, categorization. Just like with my monthly bill example, this is a really direct type of coding. In this case, an entire edge of the

card was taken up to represent categories. Each card could belong to multiple categories, so from this alone, Doug was able to build up a pretty complex network of roughly related data. To keep everything coherent, Engelbart also kept special quote-unquote master code cards, which provided a label for each hole. Just one of these note decks had categories for MIT, sociology, time-sharing computers, related works, and economics, just to pluck out a few categories. From this list, we can already see something crucial to Doug's approach to the information problem. His work was very interdisciplinary, meaning that he wasn't just tackling things as an electrical engineer. This will come up over and over again with NLS, but Engelbart was looking at multiple approaches to augmenting human intellect.

Really, for a problem this big, I think that's the only way to go. Humans are messy, and any system that has to deal with us needs to take a wide array of factors into account. You could say it's just the way we think. Moving up a rung in complexity, we get to actual numeric indexing. Categories are cool and all, but this is where the power of edge-notched cards really explodes. You see, it's pretty easy to encode numbers along one of these cards' edges. There's a handful of different ways to handle this. None of the encodings that I've seen are binary, though. Engelbart in particular seemed to favor the so-called 7421 encoding scheme. I know, that may sound like a pretty complicating name, but it works out to be really simple. You have a set of holes marked M, 7, 4, 2, and 1. The value is just the sum of

the notched holes. So if you cut out 2 and 4, the value is 6. You always have to cut two holes, so if you need to only have one of the explicit values like 7, 4, 2, or 1, then you also have to cut a notch in the M hole. That's just a rule of this encoding scheme. This opens up any number of possibilities, pun intended there. It all comes down to organizing and connecting small chunks of data. One of the simple expressions of this is grouping cards. Think of it like a more complicated version of categorization. Engelbart used this in a few ways, the most relevant being to store information across multiple cards. You see, as cool as edge notched cards are, they can only hold a relatively small amount of writing on their face. It is an

open card, so you can write whatever you want, you can paste on news clippings, you can even do drawings, but there's still a limited amount of space. To get around that, some of Engelbart's note decks group cards into series, each identified by a number. The actual series number is something like 1, 2, 3, and it's notched onto an edge. Then the face of the card has its order in that series printed on it. So, using multiple cards, he was able to reconstruct larger chunks of text. The next level up in complexity is implementing data hierarchy. Doug wrote about this very explicitly in Augmenting Human Intellect. Quote, I once tried to use my cards with their separate little concept packets in the process of developing a file memo outlining the status and plans of a research project. I first developed a set of cards upon each of which I described a separate consideration, possibility, or specification about the memo.

Engelbart continues, I separated the cards into three corresponding groups, which I shall call specification, organization, and content, and began to organize each of them. I started with the specification group, it being the highest in nature, and immediately found that there were several types of notes within that group, just as there had been in the total group. Becoming immediately suspicious, I sorted through each of the other two main groups and found similar situations in each." Engelbart continues to explain how he notched hierarchy into these cards so he could have groups and subgroups separated. Now, full disclosure here, I haven't been able to dig up this exact deck of cards that is mentioned.

It's likely that these notes just weren't kept around. This is part of that ephemeral nature I was talking about. They were probably used for this memo and then trashed once done. The groups and subgroups could have been connected using a numeric index, or they could have been connected via some other notching scheme. By looking at other note decks, I'm fairly certain that Engelbart would have implemented this using a numeric index, though.

There are some cards from the Stanford collection that contain snippets of text with a numeric key, but I haven't been able to find the code card that describes them, so the best I can do is an educated guess. As near as I can tell, this would work something like the Dewey Decimal System. You choose a large group number, say 100, to represent the specification cards. Then you just organize things such that anything from 100 to 199 falls under that larger group. You can do the same with 200 to 299 for organization, and so on. By doing so, you build up cards that belong to groups and subgroups. We can pull something really important from this besides just a kind of neat approach to

organizing data. We're seeing how, even prior to writing Augmenting Human Intellect, Engelbart is looking at ways to manage relationships between ideas. How he goes about this task is informed by edge-notched cards, not just what they're capable of, but also where the technology falls apart. A little after that passage I quoted, Engelbart explains how his edge-notched approach didn't quite go far enough. In trying to do flexible structuring and restructuring with my experiment, I found that I just didn't have the means to keep track of all the kernel statement cards, and the various relationships between them that were important. At least by means that were easy enough to leave me time and thorough capacity enough for me to keep in

mind the essential nature of the memo writing process. But it was a very provocative experience, considering the possibilities that I sensed for the flexible and powerful ways in which I could apply myself to so universal a design task if I had the necessary means with which to manipulate symbol structures. End quote. A simple pack of cardstock was getting really close to what he wanted, a way to manipulate symbol structures, a way to connect ideas and manipulate those connections, but it wasn't fast enough. You have to stop working on your primary task, in this case writing a memo, and then consult your magic cards.

That pulls you out of your work, it eats up time, and it can break your train of thought. This shows us how Doug was starting to think in terms of augmentation. Whatever he built would have to be fast enough that using it wouldn't slow down a human. It had to be easy enough to use that you wouldn't have to stop and think about it. The human-machine interface, whatever that would look like, would be crucial. The final piece here, and what I think is really the coup de grace, is linking. That's right, we are still talking about hypertext. Now, as far as I'm concerned, this is the most interesting part of edge-notched cards,

and this is also something that Engelbart was using, at least in a small capacity. The setup is this. You give every card in your deck some unique numeric identifier. Start at, say, 000, and work your way up to 9 nine. For the size of cards that Engelbart was using, three digits fit pretty nicely onto one of their smaller edges. Then, when you want to reference another card, when you want to, say, link out to another idea, you just write the three digit ID down in your note. This shows up a number of times in Engelbart's note decks. So besides the cool factor, why does this matter? Why am I so excited about these weird strips of cardstock? Simple. In the 1950s, Doug Engelbart had his very own Mimex. That is, Engelbart had built up a Mimex-like system with some more personal tweaks.

He never called his pile of note decks a Mimex, this is just my own projection, but I think it really fits. The changes here come down, largely, to the difference in medium. Sure, Bush's Mimex used microfilm, but when you strip away that facade, it was really a machine that dealt in punch card data. That's not to say that Bush's machine used a worse medium, just that any choice of medium has pros and cons. For punch cards, we get a larger volume of data storage at the cost of flexibility and imposed sequential access. The core use case for the mimics was to store an entire library in the space of a desk. In that sense, links, the associative trail of thoughts, were a secondary function. They facilitated the

comprehension of pre-existing data. Engelbart's own paperized mimics had a similar array of ups and downs. In general, we're looking at something of an inversion here. Edge-notch cards came with improved flexibility. They allowed for fully random access, but had limited storage potential. Besides only being able to store small chunks of data on each card, there's also a practical limit to the overall size of a deck. If you have too many cards, you can't thread a needle through your deck. But these limitations, I think, pushed Engelbart in a crucial direction.

In addition to everything I've covered, edge-notched cards were a very personal medium. Now, this isn't a new or unique idea, look no further than as we may think. Bush conceived of the Mimex as a highly personalized machine, allowing users to explore and connect ideas in their own ways. But this all revolved around existing printed media. The idea of a self-contained library is actually a limitation, especially if you didn't write all those books yourself. Conversely, no one was publishing books on edge notched cards.

In fact, in all of my obsessive scrounging and scouring, I've only found a handful of notched card decks that came with pre-printed information. It was very much a personal filing system for your personal information. The companies that made notched cards even played into this. For a small fee, you could send out to have custom cards produced with custom labels for your own encoding schemes. Engelbart even had ads and pamphlets for these custom card services tucked away in his notes. This impacted Doug's thinking on how human intellect could be augmented. When you get down to it, his later theories and works weren't in

service of a desk-sized library. Instead, he was looking for a way to organize your own thoughts, and that had to be fully personalized. This gets us up to the elephant that's really been in the room this entire episode, augmenting human intellect, a conceptual framework. I've been pulling from it throughout, but I haven't really dug into what the paper is all about. That's because we needed to get the context and history first to understand where Doug's coming from. The basic purpose of augmenting human intellect is to describe a method for solving the information problem. As the paper's introduction explains, We do not speak of isolated, clever tricks that help in particular situations.

We refer to a way of life in an integrated domain where hunches, cut-and-try, intangibles, and the human feel for a situation usefully coexist with powerful concepts, streamlined terminology and notation, sophisticated methods, and high-powered electronic aids. End quote. That could be taken almost directly from Vannevar Bush. The introduction even gives a short description of a near-future architect using an electronic clerk to design a house. This fictitious architect even builds up a data tape describing the house and then passes that off to a co-worker. This really closely mirrors as we may think. In this, Engelbart is really wearing his greatest influence on his sleeve. He's describing a future where we have a new way to deal with complex problems and how that will change how humans operate.

to deal with complex problems and how that will change how humans operate. But from there, a rift starts to open up. You can already see the differences even in that short passage from the introduction. Augmenting Human Intellect is concerned with, as the title puts it, a conceptual framework for how humans can be improved. Most of the paper is concerned with building up this framework, and it all starts with describing the differences between us human folk and the tools we use to handle information. While Bush takes a pretty loose approach to this, Engelbart comes in with all the fine details. We're dealing with an actual scientific paper that details a scientific approach It's not really the same kind of general audience piece that Bush wrote

One of the core concepts in augmenting human intellect is the so-called H-LAM-T framework I know, that's a pretty clunky acronym and I'll try to only use it a couple of times, but we do need to go over what it means. At least, if you ever want to read Engelbart, then it's good to have this vocabulary. H is for, well, human. That's a la us. L is language, which in Doug's prescription includes internal language such as thought. A is for artifacts, computers, keyboards,

notched cards, anything that humans use to help us out. M is for methodology. Those are the methods and approaches we use to accomplish tasks. And T is training. In other words, how H learns to L-A-M. how H learns to L-A-M. This is the framework and vocabulary that Engelbart uses to explain the idea of augmentation. From this framework, Engelbart adds in one more concept, that of hierarchy. For reasons I'll get into, I think this is where some seeds for a later struggle are planted. Augmenting human intellect takes a two-fold approach. The whole conceptual framework part, the H-LAM-T stuff, is all about how humans function on their own. This looks at how us flesh folk have learned to deal with data and our limitations so far. The second fold is about how to go beyond. That's the augmentation part. This part is deeply concerned with how human processes can be extended by computers, by an improvement of the language, artifice, and

methodology, and how we can interface with these systems. So really, it all comes down to finding a middle ground between the flesh part and the machine part. At the core of everything here is data management. Now, Bush saw this in terms of trails of ideas all linked together, something like a line. Augmenting human intellect shows that Engelbart saw this as a little more detailed. For him, humans worked in terms of hierarchy. Think about a tree with branches. Personally, I think this was a fundamental misunderstanding. I subscribe more to Ted Nelson's view that humans work best in a state of something like free associations. Nelson's hypertext is a dense, confusing maze. It's a web

of connections. He describes thoughts linking to other thoughts, the line sometimes doubling back on itself or splitting in different directions. That's closer to what we use online today, and at least for me, it feels closer to how I tackle problems. Anyway, for Engelbart, there wasn't really that kind of web. Everything was a tree. To get further into the story, we just need to accept that as read. On the human side, he broke methodology down into groupings of processes and sub-processes. Every task we complete is composed of smaller tasks. It's work all the way down. To make the bridge into augmentation, he then compares this to the structure of computers. Quote, hierarchy, so that pursuit of the source of intelligence within the system would take one

down through layers of functional and physical organization that become successively more primitive. End quote. In this, Engelbart is spot on the money. Computers are built up from simple primitive circuits that, when combined and organized, lead to supremely complex machines. when combined and organized lead to supremely complex machines. The argument being set up here is that computers and humans work in a surprisingly similar way. Of all artifacts ever built by mankind, computers are the natural tool for augmenting humans. Perhaps you could put it as they're a device made in our own image. There is more to the theory part of augmenting human intellect, but what I've outlined here are the big important parts. Basically, that you can

structure everything that we flesh folk do as hierarchical. Computers are hierarchical, and that the human-machine interface will be the key in improving the human condition. In the latter part of the article, Engelbart gets into the details of how he wants to accomplish this. This is, for all practical purposes, the first hint we get of what NLS is going to look like. This part of the paper is also presented in a pretty unique way, as a demo. The entire section is structured as a casual conversation between a protagonist named Joe and the reader. Joe, who uses this proposed augmentation system, walks the reader through a demonstration of his day-to-day activities with the machine. Once again, this style is actually really reminiscent of As We May Think,

just really kicked up another notch. In Bush's paper, he describes a science fiction scenario, and Engelbart just fleshes in all the details and adds his own twist. To kick things off, Engelbart describes a wild-looking computer terminal. It has two screens mounted at a slight angle, a dangling light pin on an arm between the screens, a small corded keyset to each side. It's meant to be operated ambidextrously, either with a hand on each keyset or one hand keying while the other uses the light pin. Also note, these aren't keyboards here. A key set as described by

Engelbart is closer to a set of piano keys. You enter data by playing chords to the computer. Design aside, the weirdest part for me has to be the fictional treatment. Joe even has dialogue with the reader. This is straight up a sci-fi short story. At one point after a quick demonstration of the complex machine, Joe explains, quote, I guess you noticed that I was using unfamiliar notions, symbols, and processes to go about doing things that were even more unfamiliar to you. End quote.

I don't know about you, but it's just strange to see this kind of writing in a very scientific paper. But hey, it shows the work's lineage really clearly, and I think that's a little delightful. Getting past that, we can start to see Engelbart's version of hypertext taking shape. The first core tenet is flexible text editing. I know, sounds boring and basic, but this was still early days for this kind of technology. Engelbart describes a full-screen text editor that featured cut, copy, paste, and undo. Formatting would be handled automatically or manually, and you would even be able to search

and replace strings. This is all second nature now, but at the time it was revolutionary. Most people were still entering data into computers by punching cards and feeding them into a reader. The text editor was really placed at the center of this human-machine interface. The text editor was really placed at the center of this human-machine interface. Outside the more mundane features, we get the interesting stuff. Engelbart, or rather, the fictitious Joe, describes a simple dictionary system where words could be highlighted and then searched for in a digital file. The two screens came in handy here,

since a user would be able to search definitions while still viewing their original text. The whole idea here was to keep you in the zone, so to speak, at least as much as possible. Further facilitating this man-machine interface was a shorthand system, a digital stenography of sorts. As Joe describes, this theoretical computer system allowed users to, on the fly, define abbreviations. This lets you build up a dictionary of mappings from abbreviations to full words or phrases. You can then switch between showing what you actually typed in or viewing expanded abbreviations. That is, you can switch between different modes. This is another core feature that we should be keeping track of. Call them modes or call them views. This becomes a

bigger part of NLS down the line. Essentially, a mode is one way of representing hypertext data on the screen. The abbreviation example is probably the most simple use case, but we can also look at it in terms of, say, a nested list. One mode will show just the topmost elements of that list, while another mode will expand and display all sub-elements. Modality here is really a way to get more out of your data. The flexibility of choosing what you see or how you see it means you can view and edit complex data on the same screen. We should all be familiar with this from the modern internet. You can either view a page as its raw HTML or as a rendered page. Each view has its own purpose, but is representing the same fundamental data. This type of approach

also meshes well with Engelbart's view on hierarchy. The next big feature in Joe's theoretical machine is, of course, the link. Engelbart describes links in a kind of clunky way, explaining linking statements with descendants and antecedents. I don't think we can really blame him. Today, you can just say, oh, it's a link, and everyone will get it. But in 1962, you couldn't even really say computer without some explanation. What's unique here is that not only can links be one way, they can also be bidirectional. That is to say that each link contains extra information

about a relationship between two pieces of information. And once again, this is where modes or views come into play. By changing your mode, you could view, say, just text that links to a point. Or you could flip it around and view everything that links to a certain chunk of text. or you could flip it around and view everything that links to a certain chunk of text. The big idea here is that via this machine, thoughts can be connected and structured, usually in some hierarchical form, and then you can traverse and manipulate those structures to see how your thoughts work. This should sound somewhat similar to what we already covered. Rearranging ideas, connecting them, and pulling out specific ideas, that's exactly what Engelbart was doing with his edge-notched note decks. The jump here

is taking that basic process and souping it up with a computer. The final piece that I want to highlight from this fictional demo is so-called team collaboration. As the fictitious Joe puts it, all this takes is a series of identical terminals all helmed by different users. Quote, they can all work on the same symbol structure, wherever they might wish. If any two want to work simultaneously on the same material, they simply duplicate and each start reshaping his version, and later it is easy to merge their contributions. The whole team can join forces at a moment's notice to pull together on some stubborn little problem or to make a group decision. End quote. Once again, this is a radical departure from anything that was possible in 1962, and the core feature all comes down to,

was possible in 1962. And the core feature all comes down to, more or less, a really advanced text editor. So far, we've seen it described as handling stenography, links, multiple modes, and now version control and collaborative editing. These are all activities that can be done by hand with pen and paper, but would take way too much time. By offloading the busy work to a computer and carefully structuring the interface, Engelbart argues that the pace of these processes could be radically increased. Humans could work faster, and therefore could work more. The secondary effect, the piece that rides just below the surface, is that the system Engelbart describes could help humans work in totally new ways. Alright, that brings us to the end of part one of this NLS adventure. As you've no doubt noticed, we haven't even hit NLS yet, but don't worry,

that's by design. This has simply been the backdrop that I need to paint for us to really pick NLS apart next episode. As things stood in 1962, Doug Engelbart had taken, as we may think, to its next logical step. I think this step was helped along thanks to the use of edge-notched cards, a pretty esoteric technology that seems to have been tailor-made for primitive augmentation. In sum, augmenting human intellect gives us a roadmap for the challenges NLS will need to overcome. It should come as no surprise that we're getting a pretty long list of very technical challenges. prize that we're getting a pretty long list of very technical challenges. This ranges from finding a way to share computational resources between multiple users to figuring out how to actually write a full-screen text editor, and even down to some of the mundane things like funding. Next episode, we're going to see how Engelbart moves forward towards human augmentation and

look into the specifics of NLS and how it's stacked up against the systems it would inspire. Before I sign off, I have a few quick programming notes to throw in. First off, I want to put in a plug for another show, The Eastern Border. They recently ran an episode on the history of computers in the USSR, which I think is pretty cool. There's definitely some stuff in there that I might have to explore on my own a little later on. So if you like my kind of content, then I'd highly recommend checking out The Eastern Border. Also, it's just a good show that I really enjoy myself. Second, I want to jump back to the matter of edge-notched cards.

Like I mentioned in the show, these weird little cards have become a big area of interest for me. Over the last few years, I've been tracking down as much data on them as I can, and let me tell you, it really isn't that much. So, dear listener, if you or a loved one knows anything about edge notched cards, then please hit me up. I'm really looking for examples of their use in the real world, since those kinds of personal accounts aren't preserved very well. So if this is you or someone you know, please get in touch. I'd love to pick your brain. Anyway, 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 history 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, 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 for the direction of or signing up as a patron on Patreon. Patrons get early access to episodes, polls for the direction of the show, and a few bonus episodes. 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 Advent of Comp on Twitter. And as always, have a great rest of your day.