Advent of Computing - Episode 137 - Edge Notched LIVE

Episode Date: August 18, 2024

LIVE from VCF West 2024, my talk on edge notched cards! Since this is a live recording from an auditorium the audio is a little boomy, so be warned. Actually, I'm pretty sure this is the same space that CHM uses for some of their oral histories. What I have today is just the audio component. VCF will be posting a full video eventually, which I'll be sure to pass around.

Welcome back to Advent of Computing. My name is Sean Haas, and I am breaking my usual format. This episode I'm calling, I think, Edge Notched Live. It's a recording of my presentation from VCF West a few weeks ago. Now, if you've been listening to the show for a long time, this is kind of going to be a rehash of some things I've been covering for quite a while on my quest to understand early data management. However, you may also enjoy this episode. It's a consolidation of a lot of the work I've been doing towards Understanding Edge Notched Cards. It puts a lot of bits and pieces I've been pulling together into one place and makes a little nice presentable package. At least, that's what I was going for.

In that sense, it's kind of similar to the live episode that I recorded when I was down at VCF SoCal at the beginning of this year. Now, before I roll the tape, I do have a couple announcements and kind of state of the pod stuff. The state of the pod is, as usual, really well. VCF West was fantastic this year. I'm really glad I decided to go and got the chance to speak. I keep having the humbling experience of meeting listeners at these in-person events. So if you made it out, I'm so glad I got to meet you. Thank you so much. Y'all are really why I keep doing the show.

Thank you so much. Y'all are really why I keep doing the show. Also, since I have the audio recorded from VCF, I decided to take a bit of a break for the last two weeks. I haven't been working on production, which has been very nice. I've actually been able to do some retro computing stuff. I've been working on restoring some of my old systems. But that said, I am jumping back into my normal two week production schedule as of well, as of today, as I'm recording this. So there's going to be just a normal,

plain, not bite sized, not live episode coming out in two weeks, then I'm just back to the usual grind. But it has been nice to take a little bit of a break. That segues nicely into my other announcement, which is I'm going to be taking an honest-to-goodness real-life vacation in October this year, which I wasn't thinking when it was initially planned, but October traditionally is my only event month on the podcast. It's spook month. It's all Halloween-themed episodes. Perhaps you see the issue here.

I'm going to actually be out of the country on a vacation for half of October this year. I wasn't really thinking ahead. In classic Sean style, I don't plan things out very well. My big brain plan that I've come up with, which I'm going to stick to since I'm saying it out loud now, is I'm going to kind of change up my format for this spook month. Instead of doing two large episodes, I'm going to do four small episodes. Those are easier for me to produce and kind of slide in as space opens up in my normal life schedule. So this year, we're going to have four small episodes,

each on something just a little bit spooky. So it'll be classic fare, just spread out. I think that'll work with my travels and work and my normal life and production. So between now and then, if anyone has ideas for spooky episodes, I have some I'm working on already, but I'm very open to anything high strangeness or spooky or annoying around computer history. Which takes me to my final announcement, which is something very annoying. If you're a patron or use Patreon, you may have seen the news about Apple and Patreon. It's not very good. Essentially, what's going on is Apple is forcing Patreon to charge 30% extra on the iOS app.

So that means if you're a new member and you join any Patreon using iOS, you're charged an extra 30%. And they're also changing how some of the scheduling around billing works. It's all annoying and kind of confusing. So my recommendation is if you're ever considering being a patron, do not do it through the iOS app. Just do it online. That doesn't charge you an extra 30% for no real reason. So if you feel the need to join up on my Patreon, get some bonus episodes, get some early releases, do it using the website. Don't do it on iOS, please.

Now, with all of that said, I'm going to start rolling the tape. This is my talk from VCF West 2024. My name is Sean Haas. As was just announced, I'm a freelance historian. I normally host and produce the Advent of Computing podcast, where I cover the obscure and less well-documented history of computing. The reason that I tend to veer towards that is because there are large swaths of computing history that are very well covered, like the history of Apple and Microsoft. Everyone knows that. There's dozens of books.

But there's these huge amounts of the historical timeline that just aren't preserved or are sitting in archives and haven't really been looked at yet. And we're just now hitting a far enough point away from those events that we're starting to get the context and enough resources that we can actually do real history about the computing revolution. So today I'm going to be sharing with you one of those stories that I've been digging into for quite a few years now. It's the story of pre-digital hypertext and how hypertext initially was developed. Just to give the broadest of outlines of the discussion today, we're going to be going from these weird-looking card stock things that look kind of like punch cards but not really and working our way up to the first automated digital links on a computer. That means that we are talking early hypertext.

We're not going to be talking about the Internet at all today. Once again, the reason for this is that transition from early hypertext systems to the modern web, it's really well documented. That was in living memory for many people. But this earlier period, it's older, it's much less well documented, and it hasn't really been treated academically yet. So my research is primarily looking at these periods that haven't been given proper treatment. But as I said, it's early hypertext,

and that's very different than the modern web. So I want to start off by going over what some of the key features of early hypertext are just so we're in the right mindset, so we're not thinking like, where's the YouTube on this? Why can't I like these cardstock slips? Where's the YouTube on this?

Why can't I like these cardstock slips? The crucial idea that we need to understand when talking about early hypertext is the information problem. This was coined by Vannevar Bush in this paper called As We May Think in 1945, but it's something that existed by different names, I think, ever since humans had the first libraries. So there's probably someone over in Alexandria going,

oh, gee, I got an issue with this stuff. The most succinct way to describe the information problem is that too much information is the same as not having any information. We reach a point as data piles up, as sources get shelved, where a single person, unaided, can't manage all that information. You can't remember where everything's stored. The problem that leads to isn't just you can't find your favorite book, it's that progress grinds to a halt. There's an old adage that we stand on the shoulders of giants, but if you can't find what those giants were writing, you can't really get on top of their

shoulders. So that's what Vannevar Bush started seeing in the 1940s. He was the director of the Manhattan Project, and he noticed that a lot of his researchers were spending more time doing background research and struggling with that aspect of their work as opposed to actually progressing the field. Specifically, Bush envisioned hypertext as a possible solution to this, as a way to better manage information so people will be able to deal with this larger and larger pileup of data. To get into the specifics, I want to look at Vannevar Bush's solution, which comes out of that 1945 paper, and that was this theoretical machine called the MIMEX. Now, I'll be going over three core features to early hypertext.

These features are present in most early hypertext implementations. I just chose the MIMEX because it's one thing, it's pretty self-contained, and it's pretty well documented. So this gives us an easy jumping off point without getting into my wild speculation immediately. The general idea of the Mimex is that it's this desk that automatically stores books. It's envisioned as this thing

that would store an entire library just in a normal size desk, very unassuming. What's cool about it and what makes it hypertext is that instead of just sequentially looking through data, you'd be able to pull up a page of a book and link that to another page, or pull up an idea that you've loaded into your desktop library and link that to another idea. By doing that, you can construct these trails of thought where you'd be able to jump from one idea to the next to the next. And the idea is you're connecting data as associations

instead of in meaningless sequences. The first hallmark of all early hypertech systems is they adopt new technology. The Mimics in 1945, it's not built, but Bush designs it around what else but microfilm. At the time, that was a wildly futuristic technology. Bush had actually worked with microfilm prior to World War II in selection machines, the idea being he could rapidly select specific slides from microfilm. So he knew it was just on the edge of possibility to have automatic linking between slides of microfilm. The second feature is freedom.

And this is a bit free as in free beer, but free as in freedom is the better adage here. This is based off microfilm. That means that your data set is. This is based off microfilm. That means that your data set is anything you can store on microfilm. There's no concept of data encoding because each idea that you link together is just a slide. That means you can have scans of a book, you can have photos, you can have handwritten notes, you can even have scans of books where you overwrite handwritten notes on top. Part of the design of the Mimix is this flat platen where you could scan a book or you could just take a stylus and write down some notes and then load that into the system using wild new

technology like dry photography. And the third hallmark, and this is the weirdest to think about in the modern context, is that hypertext is highly personal. I don't mean that as in you have the background you want or the nice CSS styling that looks pretty on your MySpace page. I mean this was envisioned as a deeply, deeply personal system. You start with nothing. In the case of Memex, you start with an encyclopedia or maybe a dictionary, but you don't have any hypertext. You don't have any connections. You, as the individual user, have to go in and make those connections.

It's personal, but you have to personalize it. That means that there's this weird learning curve with hypertext systems. It's not something you sit down and you're ready to use. A lot of other systems like NLS, you boot up and you are literally greeted with a blank screen. You're expected to use the tools that hypertext gives you to personalize your data set. The reasoning here being that this isn't some one-size-fits-all system. It's this augmentative system. It's supposed to help with your thought process and help you manage information. So in that way, it's much more personal than something like the modern

day internet. Now, that's just a quick rundown of what I think are the core features that make early hypertext really unique compared to modern hypertext, but also for the time revolutionary. really unique compared to modern hypertext, but also, for the time, revolutionary. But whenever I see nice little lists, nice little descriptions of something that's cool, I always think, is this unique? Was this the first thing to put together free-form data, high personalization,

and really adopt new forms of technology to those ends? And the answer is no. The MIMEX and later hypertext systems aren't unique in this regard. There's actually a very rich tradition of super similar systems in what I think is the most exciting field in the world. It's in accounting and insurance. Where else do you find high technology, am I right? All jokes aside, though, banking, accounting, insurance, and some isolated government tasks like census bureaus were the first places to really feel the information problem in the modern era.

And if you step back, that makes sense. These are the industries where you actually do have a wild amount of data that on a day-to-day basis, you have to be able to manage somehow. What's really cool here is there is a rich history of attempts at managing information in better ways in these fields. This starts out in around 1890 to 1919. So there's this 20-year kind of fuzzy time period where people in accounting and insurance, mainly, are designing these automated systems for file management

and for data separation and even for data selection. I want to go through a few of those because this is where we start to see the earliest roots of what might be a glimmer of hypertext. The first is the rejection index. I have that in big scare quotes because this is an anachronism I've made up. There's no trade journals

where people in the 1890s are talking about, oh, in my accounting practice, I've implemented this new rejection index after the recommendation of my associates. are talking about, oh, in my accounting practice, I've implemented this new rejection index after the recommendation of my associates. We know about these primarily through a few letters and a handful of patents. So I've just kind of applied this rejection index name as a way to group them all together after the fact. These are very simple systems where you'd have a normal filing cabinet where you would have compartments that had a

series of these notches on the bottom, little bumps. In this specific patent, they have a series of cabinets where each cabinet has different compartments. And as you can see, they have these little bumps in the bottom, and each compartment's unique. You'd pair those with note cards that had matching notches on the bottom. How this would work in practice is if you tried to put a card with notches in a slot that didn't match up with it, the filing cabinet would reject that card. You'd be physically unable to misfile a card. In practice, this would have been used for things like keeping track of bills or keeping track of invoices.

There's these great examples in patent filings where people have these 12-drawer filing cabinets, one for each month and a series of notching unique to each month. So if you are going in to file a February card in a January slot, which no one would ever do that, but let's say hypothetically it's a bad day. You couldn't do that. Your filing cabinet would prevent you. So in that way, one of the big problems with the information problem is mismanaging data. Humans make errors all the time,

and this is a way of automating away some of that human error. The other aspect to this are patents for full selection. This in particular is something that is really exciting to me. There are a lot of different ways people did this. The earliest ones and some of the most, well, it wouldn't work in real life, but it's a good try, are these tabbed cards. So this patent is from 1896, so pretty early days. How these work is a card would have a series of tabs on the top

where the tabs would have holes in the middle. Now, there's no information on, would you get a card and then cut off the tabs you don't want? Would you apply tabs above it? How would you deal with registry? So there's some feasibility issues. But what's neat about this is it allowed you to store unorganized cards in a drawer. Then when you want to make a selection, instead of having to go to one of 12 drawers, you'd open your drawer,

you take a selection needle, a glorified dowel, and you'd find the tab you want and pull up. This, once again, invoices are on the mind, so a lot of these patents show these being used to group invoices by month. So in that example, you could keep your invoices unorganized, and then if you wanted to check January's invoice numbers, you just pull open your drawer, slide your selector through the January notch and pull up, and you have all your cards.

You don't have to sort. You don't have to file. The actual operations you're doing are automated away. There's also just some really weird patents where people are trying to kind of style on the technology and put it together. This Hargrave patent is one of the most atrocious of the lot. It has a combination of rejection indexes on the top, selection on two of the sides,

an inverse index on one side, so you could use custom filing cabinets with all kinds of sliders and drawers. But eventually, they converge in the 1920s. That's when we start seeing something that's more familiar, at least something that doesn't end in a dead end. This card from Perkins is the perfect example. It's about the same dimensions as a punch card. It also has a registry cut on one corner, much like a punch card. That would have been used to make sure that when you have a stack of cards, they're all facing the same way. Note that in the middle, though, there's nothing. It's just a blank piece of paper. It's just a little note card. You can put anything you want on that. Around the edge is a uniform series of perforations. And you'll also note that some of them have been cut into notches. That's how information is encoded on these cards, which we'll get to in a minute. The other thing I want to point out before we move on

is certain segments of those holes in this patent diagram have been broken out into fields. That's going to become important in just a minute here. Now, throughout the 1920s, companies start licensing some of these patents and creating what are known as edge-notched cards. That name kind of gets applied colloquially. These are barely related to punch cards. They come out of this totally separate filing tradition and not from this number-crunching tradition.

They also only encode metadata, and it's stored on the edge. So there's no data encoding on the face of the card. You're keeping information about your information by encoding these notches and intact perforations on the edge of the card. This is, once again, substantially different from punch cards because there's no data encoding. You're just keeping track of little, just a few little bits of information. These are also fully manually operated. There are no electronic machines to help you manage your notch cards. There are fancy machines for making these, but they're essentially these kind of insanity drill presses. I've seen some photos in collections. They're these drill presses that have

geared arrangements so they can bring down multiple drill bits at once to bore out the holes in a stack of punch cards. It's wild technology, but outside of that, if you're just using these day-to-day, there's nothing electronic about it. With just this setup, just these simple pieces of cardstock, you can sort, index, reference, and select note cards. And something else that is kind of interesting and also will become a problem is there's no standards around these. The only standard is roughly the shape and the dot pitch,

but beyond that, you can encode your metadata however you want. You can put anything you want on the face of the card. I've seen some cards in collections where people have even taped Polaroids onto their Notch cards. So it makes for this really interesting medium that has a wild amount of expressive capability. Now, I just want to drive home a couple of these points.

The first is, these aren't punch cards. The reason I want to drive this home is when I first started researching edge notched cards, I assumed, oh, it's card stock. There's some data on it. It's about the same time that Holorith starts making punch cards, early notched cards appear. So they're probably related, right?

After a large amount of patent surveying and hitting the sources, they're not. And part of that is just they're ideologically totally different tools. Punch cards encode data. Edge-notch cards don't. They only encode metadata. Punch cards store data on their face as encoded information, which isn't even a concept in edge-notch cards.

The big thing is that punch cards are used to do data reduction tasks. These solve the information problem, but they do that by taking a large amount of information and turning it into something digestible to a smaller amount of information. When you do a tabulation, you're taking hundreds of thousands of cards to get an answer out of them. You're not plucking out just one card to check. They also require machines. If you don't have a tabulator, you can't do a tabulation. You can't just grab a stack of punch

cards and go, oh yes, I can scan through these really quickly and the answer you're looking for is five. That doesn't work out very easily by hand. Because of that, punch cards rely on standards. You have to have a card that's a very specific shape, has very specific punching, or you can't use all the automated machinery. So ideologically, punch cards and edge notch cards are in completely different worlds. The other point I want to really drive home

is when I said these are manual, I really mean it. The automation here is handled by two tools. The top tool is this beautiful little hand punch. It's similar to what you might have seen on a train back in the day. But you don't actually need to use that. This would have been used to turn intact perforations into notches. But you can also do that with scissors. In the historical record, there's a lot of accounts of people just using scissors. I actually don't. I'd be curious to see sales figures on hand punches.

The other tool you need is a selection needle. If you went to an official supplier like McBee or Copeland-Chatterson, you can get these beautiful Bakelite-handled needles. They're made out of spring steel. There's a little set screw so you can change the length. They even have a logo embossing gold on the handle. Or you can use a knitting needle. You could use a shish kebab skewer. I've actually used toothpicks for some small decks. So really, you don't need to invest in hardware. The only thing you need to invest in are the cards. And once you have those cards, you have everything you need to get going.

Data here is encoded in something very similar to binary. As far as like ones and zeros, trues and falses, it is binary, but some of the encoding details are wacky. The simplest encoding schema is a dichotomous key, where you just have, you choose a position and you set a hole to be a false and a notch to be true. So if you're doing invoicing in this, which would be the most common application by and large from everything I found, you might have a position that represents paid versus non-paid invoices. As invoices are paid, you pull it out of the stack, you take your little hand punch or a pair of scissors, and you turn the invoice position from a hole to a notch.

Then come up into the month, you need to see which invoices haven't been paid, so you can put those into a pile of delinquent invoices. You'd take your pack of invoice cards, pull out your knitting needle, or maybe clean off the skewer you were just having some lunch off of, and you go to the paid position, you slide it through and pull up. Any card that has an intact hole in that position stays on that needle and is removed from the pack. If it has a notch, it falls off the needle and stays in the lower pack. By doing this, you can make a selection from an unordered set of cards. You don't have to file these cards. They can be in any order you want when you do a selection operation.

There's also a really simple algorithm. It's actually O of 1 as far as complexity goes if you use big O notation. So for any amount of cards, if you're doing a selection on a dichotomous key, that's one operation. It will get more complicated, and there are limitations as far as how big your skewer can be and how many cards you can hold in your hand. But in general, this is a wildly simple selection method.

So at this point, we can start to pull together some ideas here. I'm hoping, if I've done my job, it should be clear that there are some features here that are similar to early hypertext. This is pretty cutting-edge technology for the time. At this point, there wasn't really another way. We're talking 1920s, 1930s. There wasn't really another way to automatically index information like this. It's also very freeform because your data is just a piece of paper. You can put anything you want there. You can paste on news clippings. You can

do little drawings or even circuit diagrams. And they're also known for standards. So it has that nice dual factor of personalization. It's highly personalizable, and you also have to personalize it to use it. So is this actually hypertext? It has all these features, but where's the evidence that edge-notch cards are hypertext? That gets to the issue with this specific research, or rather, issues. There are many. Here's the two worst ones.

There are many smaller ones I can tell you about later if you catch me around the fair. The first is edge-notch cards kind of just aren't preserved very well. Now, this is speculation, but I believe a lot of them are thrown out. As I keep saying, these are used in banking and accounting and insurance, all these back office applications. That's what they're primarily marketed towards. It's a way to modernize your office. Are you going to keep around your tax records for 60 years and donate it to your old alma mater's archives? Probably not. So because of that, once again, I can't prove a negative, but I think a lot of these were just thrown out. Also, when you factor in things like medical records or grant records that were stored on here that might have implications for personal information, there may have been actual policies

in place to destroy records after a certain number of years. And these are just back office records. Edge Notch cards are also really easy to get rid of. Since they're super cheap, you can use them and get rid of them and forget about them. I've read some fascinating papers. There's actually one in the Smithsonian archive that are about naturalists using edge-notched cards for really quick and dirty data reduction. There's this one where a naturalist took notes on edge-notched cards, punched them up when

he got back to his office, and then used them to figure out how many frogs versus toads he saw in the field. He didn't keep those notes, but we have a record that he used edge and notch cards when he was doing his research. So because of this ad hoc nature, and once again, the buy-in is so low, it's a pair of scissors and a knitting needle, it's easy to get rid of them. The other thing is that archivists kind of don't know what to do with these, and this isn't a is that archivists kind of don't know what to do with these, and this isn't a knock against archivists. People who work in archives are the most important people in human

history. I love every one of them. If they're listening, you're the best. I'll buy you a beer, I'll do anything you want, but they don't really know what to do with edge-notch cards. They're not equipped for it. It's not a medium that makes a lot of sense in a traditional archive. It's not a book. It's not loose-leaf paper. It's not a punch card. It's just kind of this weird little piece of cardstock that has perforations on the edge that gets hooked on other things. It's easy to miss file. So I've talked to a lot of archivists, and they've been like, oh yeah, that might be in the collection, but we don't index by that. We don't have any information about if we have these or not. So this makes it really hard to just track these down.

Which leads to a bit of a compounding issue. So as I said, these are cheap. There's almost no upfront cost and they're super flexible. You can mix and match this handful of tools to do anything you want. It's a highly personalized system, which means it's a very ephemeral system. If someone's using this for all their personal notes and then they die, well, it's very personal. You're not really going to be able to figure out what they were doing with it

unless you have a whole lot of information up front. And that's problem two. These are very personal. There's no standard. Did I say that? There's no standard at all. That's been the bane of my existence in this research. These are just a few of the random encoding schemes

you could get printed on prepared edge notched cars. We have ways to do alphanumeric entry, ways to do different number encoding using different base sequences. One of those is a dual row of perforations. I've seen cards that go up to four. I've seen cards that have regions of four rows of perforations

and then regions of one. The sky's the limit, which means you can have really rich data sets. You can have really interesting metadata encoding, but it's also, how do you start to document this? To understand a set of edge notched cards, you need a full set of cards for just the context, but you also need the context of how they were used. You need the keys for what different fields mean. So if you hand me one edge notched card, thanks. That's not very useful. You need all this added information to

actually understand how they're being used. Luckily, there actually are some preserved sets that have really, really good provenance. One of them I was able to find with the help of Christina Engelbart, Doug Engelbart's daughter, and the Smithsonian Archive. It's actually just on the other side of the highway, a few miles away. To explain this, we need to go back to 1962. In that year, Doug Engelbart publishes Augmenting Human Intellect.

This is one of the foundational papers in modern computing. It explains totally new concepts in how to handle user interface design and how to structure data. It leads to NLS, the online system, and the mother of all demos. This leads to the computer mouse. It leads to Windows. It leads to cursors. It leads to video conferencing. It leads to all of these wild modern things. And it also leads to hypertext. to all of these wild modern things, and it also leads to hypertext. Now the paper itself is interesting, but my favorite part are the appendices. In there, Engelbart goes through how and why he arrived at his conclusions. And my favorite appendix is appendix three.

I know, anyone else love appendix three out there? Any big Appendix 3 heads? I like this because he talks about his note system. He talks about day-to-day how he wrote Augmenting Human Intellect. And he says specifically that while he was working on the paper, he kept notes on edge-notch cards. His edge-notch card system. Right after this quote in Section 3, he goes on to explain a system of linking and stitching that he developed on edge-notch cards. And we're not talking about computers here. This is before he was working in a mainframe room. He developed a

hypertext system on the pieces of cardstock we've been talking about. And the wild thing is, it's preserved. The full set of notes is actually available in Stanford's special collection. It was a lot of work to find. When you're going to special collections, you're given a finding guide. It's just a PDF that has maybe one sentence, maybe one word about what's in each box and what's in

each little file folder. They never say edge notched cards are here. They'll say something really useful like notes 1954 to 1958 or correspondence comma notes period. So you just kind of have correspondence comma notes period. So you just kind of have to look and from your context guess and hope you're right. That was the case with these cards. They were just in a big box. It was labeled as notes. But I was able to track these down in 2019. And what's crucial here is we have the context of how these were being used. That was written and published in Augmenting Human Intellect. And we also have the keys. So these two cards, well, this is just an example of what the cards look like. But this larger card with the lines on it, that's actually the schema that Engelbart was using in one series of notes. That's all the information we need to

make sense, to make heads and tails of the data set. You'll notice that there's all those single words on lines. Those are dichotomous keys. He has different categories like facts or speculative ideas. You'll also notice that he has a large set of fields there allocated for numbers. You'll also notice that he has a large set of fields there allocated for numbers. Those are especially crucial because in his note collections, each card had a unique identifier. Maybe we can see where this is going. Engelbart implemented hypertext in his edge-notch card notes by using those unique identifiers. Each card has a serial number. A link is just a number that references that serial number. You can follow that link using

a knitting needle. So in practice, he'd say, oh, you know, I have this note which reminds me about this funding statement that I read or this grant that is available that I have in this other correspondence that I took note of in this other card card number 826 or 123, and he'd just write the number down. And then when he's going back through his notes, he goes, oh, I remember there was something about a funding statement. Oh, it's in card 826. Okay, well, pull out the needle and you can pull it out of the pack. The selection for a serial number like this is a little more complicated than just a normal dichotomous key, but you can do it in a few operations.

There's also tricks you can do with how you handle your serial numbers to make those selections easier. The one just interesting thing to note that I won't bore you with the details, but 7421 encoding, this binary style encoding, is actually sortable using just a few operations.

I think it's like three needle operations can sort a pack if you're encoding your numbers using that base. Now, in practice, that's what a link looks like. That is a physical link that was on one of Engelbart's notes. It's just a number in quotation, well, in parentheses. That references the unique ID number that would be encoded on every card in the set. So in this case, he could pull up card 300

with just a handful of operations. The other feature, and this once again gets back to old hypertext being different than new hypertext, is stitching. This is something the web doesn't really do these days, which honestly I wish it did. The idea is a stitch allows you, using the same idea of this linked data, to combine pieces of information into a larger idea. Specifically how it's implemented in Engelbart's notes is he'd take that serial number and then add an order number.

So you could also look at it as like a group and series number. So here, just as an example, card 826 has parts 1, 2, and 3. And all you have to do is expand to have one more numeric field, and then you can do this kind of grouping. In practice, there's what it looks like. And I really love these cards as an example, because you can tell exactly what's going on. This is card number 129, part one and two. You can see that he ran out of space typing up on one card. He has these little overs on a lot of his cards, but you flip it over, there's more, he runs out of space again.

So he has this other card that has the same ID number and has a little sequence number added on where he finishes the idea. The idea with this, and what's so cool, is he's able to pull out a specific idea and then reorganize it, put it back in order and reconstruct the entire chain of thought. It's really interesting to note, too, that he's really making use of the medium because he hits this point where he runs out of space and instead of just going, oh'll abbreviate it i'll remember later it'll be fine he goes no i'll i have the tools to extend the data i can keep my idea going i can keep rolling with this because the system gives him all the tools he needs using this hypertext ideology so to pull this all together edge notch cards not only have similar features to early hypertext,

they also have almost the same implementation details. In the modern web, how do you access a web page? You have a unique identifier. When you reference it, you reference it by that unique identifier. I mean, a few numbers is a lot less unique than a full URL, but it's serving the same purpose. Edge-notched cards also show up at the dawn of digital hypertext. Doug Engelbart is the first person to implement hypertext on a computer. And as he's designing that system, as he's planning how he's going to implement hypertext, he's doing that on edge-notched

cards. That is a direct connection. They are, if you'll excuse the joke, the missing link to the story of early hypertext. But this is only part of the story. Edge-notched cards aren't just this weird hypertext technology that appear and disappear. They continue being used into the 1970s, if you can believe it. I actually have, and it's really weird. So I live in a pretty small town in pretty rural Northern California. Going through court documents a while back,

I ran into this price list. It has a breakdown of things that the cops in town bought in 1977. And one thing they bought was a crime database on edge-notched cards and a training course on how to use them. The cards have since been destroyed, but there's records of these being used at least as late as 1977 and just out in the middle of nowhere. So these, I personally think, it's secret backroom technology. It's something that's used in the middle of nowhere. So these, I personally think, it's secret backroom technology. It's something that's used in the back of an office

that people don't really talk about, but it's used for decades. This opens up a lot of questions. Was Engelbart the only person using these for hypertext systems? Was he the only person writing these automated links? What else were people using these for? We know they're used for hypertext, they were used for ad hoc data reduction and survey work,

and they were used for office records. But what else is going on? This is just an entire hidden field that we don't have eyes on yet. Currently, there are two fronts to this research. The first is adding depth. Now, this isn't me, but I'm very excited about it. I have a contact in New Zealand whose father recently passed. When he was going through his estate, he found a few filing cabinets full of edge-notched cards. His father was a naturalist.

He was using them to keep his field notes. My contact is currently digitizing those cards. Actually if what he says is to be believed, they should be digitized right about now. What's really exciting about this, besides having a new data point, is he's writing software to do full image analysis to extract the indexes and preserve photos of the cards. And I can't wait to get my hands on that software. That's going to allow me to take the scans I have, all the, well, all the six full sets of edge-notched cards I've been able to find, I have those digitized. Using this software, I'll be able

to pull out those indexes, store that data in a database, and actually operate on them in a computer instead of manually. Because these are old, they're crumbling. I don't really want to pull out a knitting needle and go to town on these. If I damage them, that's something that's lost forever. Also, archives usually don't let you bring pins inside when you're going through the special collection. I don't know how an archive would feel if I showed up with an ice pick and was like, hey, I want to look at box 387 from 1958, please. Can you bring that here?

No, just for a minute. It'll be fine. I'm not going to do anything bad. So I'm really excited to have a way to use this on a computer. I think that'll give us a lot more insight into practically day-to-day how these could be used. And the other side to this is adding breadth. And this is the end of my presentation and my call to action. Have you or a loved one seen these cards? Back in the 70s, did you have a week at work that was ruined dealing with these? Do you know someone that manufactured these, someone that threw these out? If you have, please get in touch with me. I'm going to be here all weekend. I have business cards. You can email me. I'm trying my hardest to find more of these cards because there's so few data points about them. The last academic work, well, academic, the last industry work on edge-notch cards was

published in 1954. That means there's this huge gap where we just, we don't have a centralized understanding of what they were used for. I think in the future, in the coming years, as I collect more information about these, as we start digitizing cards, as we start using them on computers, as we start finding more cards, there we start using them on computers, as we start finding more cards. There's going to be a really fascinating story here. And as we've seen with the connection to hypertext, we're just starting to scratch away at the surface of that. Thank you so much for listening to my talk. If you want to hear more of my ramblings, I host Adren of Computing. It's a podcast. You can get it anywhere anywhere and if you've seen these cards

please come find me I will reward you handsomely I'll at least give you a firm handshake and a thank you applause applause applause applause do you want to take any

questions? I would love to take questions alright so if you have questions, please raise your hand. I'll come to you just so that it's recorded. Hi, Sean. Thanks very much for the talk. It was wonderful. I'm a big fan of the podcast.

Thanks. While doing some research on actual punch cards, this is related but not exactly about edge cards, I found it very difficult to find any sources that would describe how you actually did work using a stack of cards. I mean, I have a stack of cards as sales contracts. Yeah.

Why they didn't turn that into something useful? The best source is this. I actually have little citations to it everywhere since almost all my images came from here. There's this book from the first edition was in 51, the second in 54, I think. It's called Punch Cards, Their Application to Science and Industry. It's a collection of publications from trade journals and academic papers that's on edge-notch cards and punch cards. It's scanned on the Internet Archive for free. I highly recommend that. I have a personal copy that I think I've dog-eared to oblivion.

But that has full explanations of day-to-day how people are using them. Thank you. Hello. I'm fascinated not only about learning the system, but how far it propagated throughout the world. You mentioned that you've heard about collections of these cards in New Zealand. Yeah. I was wondering how did this system get propagated?

Did he present papers, presentations? I guess many of us know about the infamous Mother of All Demos. Did he do presentations about this particular system and academic papers? Oh, you just mentioned some. That's a really good question. So Engelbart didn't present about Notch Cards. They do show up in Augmenting Human Intellect, like I said. So that was part of the propagation. There are also just ads. There were a lot of companies that manufactured edge notch cards. One of the biggest was McBee, which later becomes Royal McBee.

They ran ads everywhere. You can still find a lot of images of, they're really campy. They'll have like photos of someone in a business suit standing astride the world holding a nosh car being like simplify your business solutions conquer the world um so that was a big part of the spread there's also a lot of academic disciplines chemists used these a lot there's actually in punch cards or applications science and industry there's some fascinating papers about how chemists were encoding information about compound structures on the indexes in notch cards they also i think some of the spread was people doing research in different areas of the world there were um i'm gonna mess up the details since I haven't read the paper in a while, but there were some TB studies carried out in India during the British occupation that used

edge notch cards. And they were using, I think, Copland-Chatterson cards, which were manufactured in England. So they're spread via that vector, but in general, they spread by industry goals, kind of. So the cards that were found in New Zealand, have you seen any of those images yet, or are you still waiting for the entire project to be scanned? I've only seen little phone camera images so far. I'm waiting for the full scans. I know that the code's going to be on GitHub,

and he was saying that he's donating the full set to a natural history museum. But I'm still waiting on the full photos. So the reason I ask is similar to the previous question. Were they significantly different from cards you've seen in the U.S.? No. So that's what's really interesting. And this is one of the open questions right now that I need to go on a couple trips to a few archives to finalize this.

I need to go on a couple trips to a few archives to finalize this, but most cards, if they have that registry cut, if they're related to this patent, so I think people were licensing this Perkins patent, and some have these circular cuts at the top. There are some that have a straight cut at the top that I think weren't under license. But in general, around the world, they all look very similar. The form factor ends up being a little more square than

the earlier patents. But in general, they all look roughly the same. That was kind of a rambling answer. I hope that answers the question, though. Hi, thanks for your talk. In terms of early paper-based real hypertext, would you consider the card systems of Paul O'Leary in Belgium in the 1800s to be or to not be hypertext? My man, 100%. I personally think if they're not real hypertext, they're really, really close.

Because he was using, I guess, for those who don't know, he was using a very much more sophisticated indexing system similar to the Dewey Decimal System, but was able to encode much more information about data. And that was used for referencing and connecting data, which is, that's really hypertext-y. So yeah, I consider it hypertext, at least. Thank you.

Edge-notched cards sound awesome. Do you know can you still buy them? I wish. You can't buy new ones. They show up from time to time on eBay. You can get them there by the one and two. I've actually talked to a couple of people that are trying to manufacture some just for personal use and it's not very easy since you do have to have registry

and I've seen machines that are still extant for cutting these out but I don't know if they work so as far as buying new ones no, best you can hope for is new old stock. It looks like most of the usage was predominantly on business and academia. Was there any effort at all for people to use this in a home or an entertainment setting? I've read, so I, you could probably use these for a choose-your-own-adventure novel. I haven't ever seen them used, well, that's not entirely true. There were a couple of sets that were sold as

field guides, so there's like an edge-notch guide to birding or an edge-notch guide to rocks that I found, where it was for personal use. As far as pushing people to use these as a personal filing system, the closest I've seen is there's this really slim book I tracked down that was, I think it's just called an introduction to personal filing, that was trying to advocate for these as for keeping your personal taxes. But I haven't, excuse me, I haven't seen much about people using these for like journaling or in the home

besides small time business stuff. Any other questions? Looks like that's it. Thank you very much. Thank you. Thank you very much. Thank you.