Advent of Computing - Episode 26 - Memex and Hyperlinks

Episode Date: March 22, 2020

The widespread use of the internet has shaped our world, it's hard do imagine the modern day without it. One of the biggest featured would have to be the hyperlink. But despite the modern net feeling so new, links actually date back as far as the 1930s and the creation of the Memex: a machine that was never built but would influence coming generations of dreamers. Like the show? Then why not head over and support me on Patreon. Perks include early access to future episodes, and stickers: https://www.patreon.com/adventofcomputing Important dates in this episode: 1927: Differential Analyzer Built at MIT 1938: Rapid Selector Built by Vannevar Bush 1945: As We May Think Published

If I had to pick a tagline for this podcast, I think it would be something along the lines of the internet isn't that new, or maybe there was no birth of the internet. Now, you may think that those both sound a little sensational, and you'd probably be right. But the fact remains that both of those are totally true. Let me explain. The internet is a really good idea. In fact, it may be one of humankind's best ideas.

It's able to connect everyone around the world, warehouse the sum total of all human knowledge, and it makes all that information easily accessible from anywhere. If you step back and look at the network in total, you can pretty quickly see how ambitious the whole undertaking is. It's simply too good of an idea and too big of a creation for any one person or even a group of people to have created. Each tiny facet of the internet has its own rich history of innovation, and collectively,

all of those inventions lead to the modern day net. One of those facets that has become key to the modern internet is the hyperlink. Those are the little underlined words that you click on to go from one page to another. It's one of the core features of the internet, and its story beautifully illustrates my point. You see, the hyperlink goes all the way back to the 1930s and a machine called the Mimex. Now, Mimex wasn't a computer. It didn't connect to any network. Networks didn't even exist yet. And it was never actually built. Despite all of those handicaps, it served to spread the idea, starting a chain of thoughts that would eventually link up with the World Wide Web we know today.

with the World Wide Web we know today. Welcome back to Advent of Computing. I'm your host, Sean Haas, and this is episode 26, Mimex and the Hyperlink. Today, we're dipping back into the long history of the internet. But instead of looking at the purely technical side, we're going to examine some of the roots of the ideas that would lead to the modern network. And to do that justice, we need to go back to the 1930s and examine the Mimex, a machine that was designed to implement something very similar to the internet, but never actually materialized. You could dismiss it as vaporware, or you could call it a contagious

idea. And I hope that I can convince you that the latter is a better descriptor. At its core, Mimex was a machine designed to store and organize large amounts of information. Data could be added to Mimex, edited, sorted, and distributed to other Mimex users. And here's where I think it gets really interesting. Mimex has one of the earliest descriptions of hyperlinks. Each entry in Mimex could be linked to other entries, creating a chain of links much like we see in the internet today. On its own, that would be an interesting enough story. But there's a lot more to Mimex than just a modern-sounding design

that was put together before electronic computers. The context around its creation and its creator make Mimex all the more interesting. And like the later ARPANET, Mimex even has connections to the US government and, believe it or not, the atomic bomb. And the most direct connection comes down to its creator. atomic bomb. And the most direct connection comes down to its creator. Vannevar Bush, the scientist who oversaw and, in part, helped to initiate the Manhattan Project, also created Mimex. But the thread runs a little deeper than that.

So today we're going to be looking at one of the earliest predecessors to the internet. What exactly was Mimex? Why was a system full of hyperlinks developed in the first place? And what would the internet have looked like if it was built in the 1940s? To understand why the MIMEX is important in the larger history of the internet, it would help to understand the perceived need behind a device like the MIMEX. Of course, no one in the 30s or 40s was going, gosh, we really need to get behind this new internet thing. But, like all innovation, Mimics was designed to solve a very specific issue

of the day, that being that there was simply too much information for any one person to comb through. This problem comes down to the fact that a single person can only do so much, and really, you can only know so much. Humans in general aren't very good at filing away data in our heads. We can't hold all that much, and we really can't recall it quickly or accurately over time. This problem has undoubtedly existed since time immemorial, but starting around the turn of the 20th century, it was appearing more and more often. Sometime in the 1940s or 50s, this concept would finally be named the information problem, but it was well known before it had a name. One of the earliest and most well-documented cases of the information problem happened during the 1880 U.S. Census,

and I think it provides a pretty good example for our purposes today. As per Article 1 of the U.S. Constitution, every 10 years there must be an enumeration of everyone living in America. Data on everyone has to be collected, compiled, and then analyzed. everyone has to be collected, compiled, and then analyzed. This is partly to collect generally useful statistics, partly for tax purposes, and most importantly, it's used for allocating seats in the U.S. House of Representatives. So every 10 years, it's important to produce an accurate and timely census. Otherwise, there could be some big problems for the U.S. government. Otherwise, there could be some big problems for the US government. Luckily, the process used to collect and compile the census had rolled along fine.

At least until we get to 1880. All told, it took 8 years to collect and compile the 1880 US census. While that's not a disaster, it is cutting it pretty close to the wire. A large part of the problem with the census just came down to the sheer scale. America had been going through a whole lot of growth during this period. So when the surveys came in, the Census Bureau had to deal with data on over 50 million people. And with the next census just around the corner, concern was beginning to mount. With the population only growing, the information problem was really starting to creep in. There were a lot of changes made in preparation for the 1890 census to try to prevent a decade-long process. And the largest area for improvement

had to be the time-taken compiling census data. You can't really do much to speed up collecting data, but processing could be improved. In 1880, this was all done unaided, simply by a small army of clerks behind desks grinding through stacks of census schedules. While this process had reduced good results, at least in the past, it was pretty slow and there wasn't much of a guarantee of accuracy. A person can only do so much with pen and paper. Some kind of help was undoubtedly needed, and it would come from a strange new machine. As it turns out, the census office would be in luck. In 1889, Herman Hallrith, himself a clerk during the previous census, patented the punch card and card tabulator. These early punch cards

were able to encode simple data as a grid of holes punched on a piece of cardstock. And the accompanying tabulation machine was able to read data from those cards, keep a running count of a few different data points, and then tell the operator where to file the card away. It's a pretty crude method, but it was just the help the census needed. Thanks to a little help from a mechanical aid, the 1890 census, despite being larger than the previous census and collecting many more data points on each resident, was completed in just six years. Preliminary data was already ready within a few months.

And thus, the information problem was conquered. Well, at least for the U.S. Census Bureau, they were able to rest a little bit easier. These types of problems, too much data for people to process, aren't uncommon in history. The 1880 census is just one example, albeit one that was solved pretty quickly. But just as the population grows over time, so too do other sources of information and data. And we don't have to look very far to find other examples. And not all exist in such a small niche as the U.S. Census Bureau. I think it's about time to introduce our main character for today's story.

Vannevar Bush was a career scientist. Born in 1890, Bush was of just the right age to have been involved with both world wars, but he'd experienced them in a very unique way. He earned a PhD in engineering jointly from MIT and Harvard just prior to the US entering World War I. Early on, Bush's career was mainly focused on radionics, but his interests would shift considerably throughout his life. By the interwar years, as early as 1927,

Bush started building a machine called a differential analyzer at MIT. We would know it today as an analog computer. Machines like Bush's differential analyzer are very crude compared to today's computers, but they're fascinating machines, and for the time period, there was nothing that approached them in terms of speed and accuracy of calculations. Over the course of the next decade, Bush improved his designs and built more capable computers. Initially, the differential analyzer and its descendants were purely used for research. But by World War II, the technology would become integral to the war effort. With digital computers only appearing near the very end of the Second World War II, the technology would become integral to the war effort.

With digital computers only appearing near the very end of the Second World War, analog systems were key in calculating ordnance trajectories, fuel consumption of fleets, and just general number crunching. These machines were designed for giving numerical answers to problems too complex or too time-consuming to be done by hand. And that specific application, they really excelled. A differential analyzer could solve equations, crunch numbers, and answer nearly any mathematical question much faster and more accurately than a human. Bush was at the core of this fast-moving field of research.

And it was while at MIT that he made a pretty important observation. At least as early as the 1930s, maybe even sooner, Bush was realizing just how much time scientists spent doing background research and referencing past studies. Running background research and digging up proper references is a very important part of the scientific method. Most innovation is possible by standing on the shoulders of giants, and those giants tend to leave behind a pretty long paper trail. The problem here was that, in Bush's approximation, it was easy to spend more time looking for past research than actually carrying out new research. This had the effect of slowing down any actual

progress a scientist could make. To make matters worse, the corpus of research would only grow over time, making the time required to dig through data that much longer. Project this into the future and the consequences should become clear. There will come a point where progress will have to come to a complete standstill. It would soon become impossible for a human, unaided at least, to sift through the ever-expanding mountain of information. Today, we have a lot of options when it comes to searching, sorting, filing, and just finding data. That's thanks largely to computers. But in the 30s, computers barely existed, and the ones

available were comparatively simple number-crunching machines. So what would an enterprising scientist have at their disposal for finding information? What would someone like Vannevar Bush be subjected to in their quest for knowledge? Sadly, options were pretty limited. The best-case scenario is that you only need some cursory information, something simple enough to get from an encyclopedia. For a search that simple, you'd only have to leaf through as few as maybe 30 books.

Entries in paper encyclopedias were usually organized alphabetically by title, and there is usually an accompanying volume that served as an index. So finding information wouldn't be that big of a struggle. But that only works for very surface-level information. A next step for deeper research would be to start looking for books on the subject of interest. So it's off to the library. And when it came to searching a library in the pre-computer era, there was one tool of choice, the cumbersome card catalog. If you've ever had the occasion to use one of them, then you know how frustrating the process can be.

In its most basic form, a card catalog is simply a large cabinet of drawers full of index cards. A card catalog could be used to store just about any kind of information, but they were most commonly used for tracking the inventory of a library. In this form, each card acted as an index for a single book, with the location of the book in the library, its title, author, and maybe a short description written on the card. To find a book, all you had to do was look through the card catalog. But that was easier said than done. Catalogs could be sorted or indexed in a few different ways. Some were sorted alphabetically by author, other by subject. Still others stored the book's title.

If you knew the author that you were looking for, and by chance the library indexed by author name, then you could pretty quickly find what you wanted. But if you didn't know exactly what you're looking for, or what information you have doesn't line up with the library's filing system, then you're in for a really long search, especially since a card catalog could get massive. That's a one-way ticket to combing through shelves of books and piles of cards. Even once a book or paper was found, things wouldn't be much better. As you dig deeper into a topic, you'll undoubtedly come across new questions, possible tangents, or adjacent paths of inquiry.

Pretty quickly, a web of new information will start to form. But for each node on that web, you have to go back to the card catalog and try to find the matching source. Throughout his career, Vannevar Bush came to know this process inside and out. It was a must in his era. Most saw this kind of work as just part of doing research, more of an annoyance than anything. But Bush saw things a little differently. He saw it as a fundamental issue with how knowledge was being handled. He had stumbled upon the information problem, and once he found it, Bush's mind got churning trying to solve it.

By 1938, Vannevar would take his first stab at trying to solve the information problem, and this is where things take an interesting twist. He identified the issue as not solely residing in indexing and document retrieval, but rather in how information was stored in general. In his mind, paper books were just as much part of the problem as card catalogs. So any solution would have to come in the form of a pretty radical shift. But despite the sweeping change that Bush envisioned, his solution would make heavy use of existing technology, and it would borrow considerably from past works. We're still in the 1930s, so any solution would have to be shaded by the technology of the time. Microfilm was one of those technologies of yesteryear that's since been totally superseded.

It would become Vannevar's medium of choice for data storage and a cornerstone of his futuristic vision. Originally developed in 1839, microfilm was a downright archaic technology by the time Bush started working with the information problem. Besides being old, microfilm is just downright simple. Essentially, microfilm is just a reel of photographic film that holds size-reduced images. A series of processes, usually called microphotography, are used to scale down an image to fit on each tiny frame on this spool of film. Oftentimes, images are reduced down to about 0.25% the size of the original. Once transferred onto microfilm, the original image is reproduced using a projector. Microfilm can be used to store any kind of data, or at least anything you can

take a picture of, but its most common use was for storing documents. Before computers and document scanners ever existed, microfilm was the best way to store and preserve a sheet of paper. The big draw here is in converting books to microfilm. Each page can scale down to less than 1 100th of its original size. An entire bookshelf can be condensed down to a spool of film that can fit in a pocket. A library could be reduced to the size of a small box full of these films. The benefits of this are immense. Instead of having to leave a lab and go search through a library, a researcher could step over to their microfilm reader. On the other end of the spectrum, a fully converted library could house an insurmountable

amount of information on microfilm, in the same space that their existing paper collection was. Bush would have already seen some of the uses of microfilm in this capacity. Its use for archival media was gaining popularity in the 20s and 30s, and in the coming years, only more data would be transferred onto microfilm. For all the world, microfilm looked like the media of the future. But on its own, microfilm was only a half measure. No matter how great your storage medium is, it's useless unless you have a convenient way to search through it. Even if all the world's knowledge did end up being transferred to microfilm, you'd have the same issues accessing

the new film library as an older paper library. The other half of the issue was finding not only a replacement for the card catalog, but a whole new way to look at indexing and searching for information. But that would turn out to be a pretty tall order. As it stood, there just weren't many alternatives to the venerable card catalog for data retrieval. Computers may have been able to fill that space eventually, but that wouldn't have been a viable option for decades to come. It would take a little ingenuity to come up with a solution. Now, I know I've been harping on the lack of options for data retrieval in this period.

The fact is, there were a few alternatives to the card catalog, but none of them were a good replacement. One option that I already mentioned today was the punch card. That technology had come a long way since its mentioned today was the punch card. That technology had come a long way since its early days at the census office. By the 1930s, Hollerith and a handful of other inventors would come to form the Computing Tabulating Recording Company, which would later transform itself into International Business Machines. As recognizable as the IBM name has become, I actually prefer the company's original name, simply for the sake of clarity. They dealt in computing, so adding machines for

the time, tabulating, or punch card machines, and recording, or accounting machines. At the core of most of this business was a radically improved punch card. These new standardized cards were composed of a 12x80 grid of holes capable of storing much more generic data than Hallrith's original census cards. In practice, that boiled down to encoding numbers and textual data, up to 80 characters on each card. up to 80 characters on each card. And with a combination of card tabulators, sorters, and punches, it was possible to produce, sort, and filter on stacks of cards. So it would have been possible to replace a card catalog with a punch card system. And in fact, some libraries in this era did supplement card catalogs with punched cards. However, punch cards don't make a perfect replacement for card catalogs.

One of the major problems comes down to practicality. Punch cards had to be punched and used via machines. Card catalogs were ubiquitous in part because they were a totally manual device. You just needed a pen and paper to add to the catalog and it was even easier to pull from. This dependency on machinery also made punch cards a lot more expensive, inordinately so compared to how dirt cheap it was to operate a card catalog. The other big problem with punch cards came down to the lack of flexibility. An 80-column card could only ever hold 80 characters of text. If used just for indexing the library, that might not be so bad. But card catalogs were used for more than that. You can't really keep notes in the form of 80

character snippets. And you definitely can't save any type of visual data that way. Don't get me wrong here. Punch cards were a really important and powerful technology. However, they couldn't do everything that a simple index card could. What Bush and his lab at MIT ended up building would be a mix of a few of these older technologies. The final device, called the MIT Rapid Selector, would use microfilm as a storage medium, and it would take major cues from punch cards for indexing data. But the blending of these two technologies would come in an interesting form. The actual microfilm reel would alternate between slides of reduced images

and slides of indexing data. If you were to hold a few of these slides up to the light, then you'd see microphotographed pages from books and studies alternated with incomprehensible patterns of light and dark rectangles. If you squint, you may have even been able to work out that those patterns match up to the holes of a punch card. The non-human readable part was where the magic of the rapid selector comes into play. Inside those slides was encoded information about the following photograph. This takes the place of an index card in a card catalog and includes details like title, author names, subject, and year of publication. The rapid selector then had an array of photoreceptors capable of reading off that pattern of dots. To find an item inside a

spool of film, you just had to take your query, convert it into the encoded patterns of dots that the rapid selector would understand, and then let encoded patterns of dots that the rapid selector would understand, and then let the machine do the rest. The rapid selector would then scan through the reel and transfer any sides that matched your parameters to a new set of film. After a few minutes, your search would be done. All in all, it makes a pretty compelling replacement for a more conventional library. Bush wasn't the first to think up this idea. There had been some stir over similar mechanisms going back to the late 1920s, but the MIT rapid selector that he built was a very early example of the technology at work.

Over the next few years, the rapid selector would be used on campuses and start to spread outside to government institutions. It was a really good way to quickly reference a library of microfilm and runoff copies. Other machines, like Bush's original, were produced, eventually making their way as far as the CIA. But not all was well with this library on tape. I think the Rabbit's Lecture is a good example of a best a best attempt technology. It could do things a card catalog couldn't, and for its time and place, it was a major improvement to conventional means of data retrieval. But the core design of the RapidSelector was majorly limited. And one of those big limitations came down to how much data the RapidSelector could process,

came down to how much data the rapid selector could process, and not just in the abstract. One invaluable source for me has been Information Selection Systems Retrieval Replica Copies. It's a wordy 1961 report from the U.S. Department of Commerce and National Standards Bureau. Besides outlining the history of the rapid selector and its use, the report gives some hard numbers on the machine's limitations. For instance, the rapid selector used by the Department of Agriculture. Pulling from the report, the machine was, quote, designed to use a 2,000-foot roll of 35-millimeter microfilm. On this, up to 72,000 frames of text or abstract material and 432,000 encoded selection entries to the material could be recorded. Searching was intended to proceed at a rate between 70,000 and 120,000

codes per minute. The search run for an entire roll would therefore take between 4 and 6 minutes. However, in this mode, only one code pattern could be searched per run. End quote. Just that passage brings up several possible issues. Firstly, a rapid selector could only handle a set amount of data at any one time. If you had a library with more than 72,000 slides worth of information, then you would have to search through multiple spools of microfilm, swapping out the film for each separate run. Add to that the limits on encoded data. Where the report says encoded selection entries, they're talking about fields like author or title that are punched out in the pattern of light and dark. When operating the machine at full speed, only one of those fields could be

searched at a time. So if you wanted to find every paper written by Sean Haas in 2020, for instance, then you'd need to run the machine at a slower speed or do two separate searches. Another problem with the RapidSelector was that, when you get down to its core, it's just a mechanical device. It may be approaching digital with its punch card-like indexing, but that's only a small part of the machine. The rest of the mechanisms for feeding microfilm, triggering photographic transfers, and the like is purely mechanical. Due to the overall complexity, a rapid selector tends to be on the delicate side, and over time it needs maintenance to avoid disaster. time it needs maintenance to avoid disaster. Ultimately, Bush's rapid selector would live as a niche solution to a larger problem, never leading to the sweeping changes in information storage he envisioned. But building the device would set Bush up for his next big leap.

The next step for Vannevar wouldn't be at MIT. Through a series of promotions and appointments, his career trajectory would only go up. In June of 1941, he was assigned as the first director of the Office of Scientific Research and Development, or OSRD. In the words of President FDR, the new department was established for the purpose of assuring adequate provisions for research on scientific and medical problems related to the national defense. End quote. So just as the U.S. entered what would be the largest conflict in human history, Bush found himself overseeing some of the most important projects to the war effort.

OSRD's portfolio was pretty diverse, ranging from developing new methods to produce antibiotics to improvements to automatic missile detonation systems. But by far, the most influential and most well-known project supervised by Bush and OSRD was the creation of the first nuclear bombs. There were a lot of names for this undertaking, but it's best known as the Manhattan Project. Started in 1942 after considerable lobbying by Bush and his colleagues, it would best known as the Manhattan Project. Started in 1942 after considerable lobbying by Bush and his colleagues, it would be one of the most ambitious scientific undertakings in that point of human history. And that meant one thing, a massive amount of research would have to be done. The lifeblood of OSRD was information, especially when it came to creating the bomb. Besides being an endeavor of

unprecedented scale, the Manhattan Project was taking place at a very pivotal time. As the last world war drew to a close, the era of digital electric computing was just beginning, and as digital computers became viable, older analog systems fell out of use. Bush had been a leading figure in the field of analog computing for most of his career, and a big advocate for computing in general. And while running OSRD, Bush was in the very nexus of the cutting edge for computing. Creating the atomic bomb required massive amounts of highly accurate calculations. Los Alamos and other related

labs ended up burning through the latest calculators and analog computers during this process. By 1944, the Harvard Mark I, a very early digital computer, would become operational. This new machine would prove invaluable during the latter stages of the Manhattan Project. And as more systems followed the Mark I, the writing on the wall was clear. Digital was going to be the future. And the future was progressing forward at lightning speed. It would seem like a logical progression for Bush to take a jump to working with digital systems. But he wasn't one to act as expected. Instead, in July of 1945, mere months before the first atomic bombs were dropped, and in fact days before the Trinity test was announced, Bush published As We May Think.

First appearing in print in the Atlantic Monthly, the article itself is only around seven pages long. But in those pages, Bush laid out something truly revolutionary. The paper is half critique of the current state of technology and half utopian view of a near future. Despite coming out at the very end of World War II, As We May Think had been brewing inside Bush's head for a while. He was already versed in the need for better information management, but while overseeing the Manhattan Project, that need came into even sharper focus. The whole first half of As We May Think is dedicated to laying out the information problem as it stood in 1945, essentially what I've covered up to this point. But Bush diverges

into an insightful take on the matter, and this is really where a truly novel solution appears. As the title suggests, Bush's article makes the argument that the reason existing information retrieval methods fall short is that they don't mesh with how humans think. People don't think of things in terms of index or filing cabinets. To quote, It operates by association. With one item in its grasp, it snaps instantly to the next that is suggested by the association of thoughts, in accordance with some intricate web of trails carried by the cells of the brain. Today, we would call these hyperlinks, or just links.

And at the time, this was a radical shift in thinking about the information problem. It basically reverses existing solutions. Instead of finding a way for people to be better at searching through a catalog, Bush is describing a way to make machines better for people searching through them. It's a matter of shaping machines in the image of the human mind instead of humans needing to adapt to use increasingly complicated systems. The mind was a source of inspiration for Bush, but he also saw that there was an opportunity to create an improved aid. To quote from As We May Think Again, quote,

It has other characteristics, of course. Trails that are not frequently followed are prone to fade. Items are not fully permanent. Memory is transitory. Yet the speed of action, A machine doesn't forget like a person, so it can make the perfect complement to a human mind. That is, if properly built. The bulk of the rest of the essay is devoted to describing just such a machine, or at least a thought experiment in creating such a machine. Bush called it the Mimex, and it's a pretty strange mix of mid-century technology and science fiction. You can think of it as retrofuturism at its best. Mimex was designed, first and foremost, to look and feel at home in a normal 1940s office,

so everything had to be built into a desk, but with a twist. At the core of Mimex, housed inside the desk, was an improved rapid selector. Bush described it as using thinner microfilm, a highly reduced image so as to be able to store more information in a smaller space. It was also scaled up from the MIT version. It was set to support multiple reels of film at once, each being able to be selected from. On the top of the desk was a collection of controls and displays. At the center of the desk were two separate screens, portrait-oriented to match the aspect ratio of a sheet of paper. This is where the pages from the internal rapid selector would be projected. But those screens served a dual purpose.

Besides just displaying each page, Bush described them as being touchscreens. They would serve as the main input for Mimics. From these dueling touchscreens, a user could call up a specific page to view and navigate through archives. On the right side of the desk was a bank of switches and buttons, essentially a supplement to the touchscreen controls. But Mimex wasn't just about viewing existing information. A large part of Bush's design was the ability to edit information and add to your library by converting pages into microfilm. On the left side of the Mimix desktop was a platen, essentially a flatbed scanner, and was used for imaging new pages. On the technical

level, this was slated to work via dry photography. It's an interesting method that can be used to copy and develop films without using many chemicals besides the pre-prepared film. films without using many chemicals besides the pre-prepared film. Dry photography was a relatively new invention when Bush was working on As We May Think, so it featured heavily in connection to the mimics. The key that made it attractive was that dry photography was, by and large, a much more simple process than traditional photography. This allowed machines like the rapid selector to go straight from an image to microfilm without using chemical baths or a dark room. The touchscreens could also be used to edit documents and add notes. These would be transferred to microfilm using the same

dry photography process. So there we have just the idea of touchscreens used for data manipulation, and that's insanely futuristic on its own, especially from a paper coming out of the 1940s. We wouldn't have anything like that for decades. But Bush went further by describing a scheme to implement the linking between documents, the aforementioned trail of thoughts in the human mind. Those would turn into hyperlinks on the modern web, and the original as conceived by Bush is pretty similar to modern links. Today, a link is made

by referencing some location on the internet, usually in the form of a URL or a website address. It's a unique identifier that a web browser can go to, pull down data, and then display. Like with the rapid selector, each frame of microfilm in Mimex was composed of encoded data and then the actual image. This made information searchable by different fields, and linking worked in a similar way to just searching. One of those fields was used as a serial number, a unique identifier for each slide of the film. To make a link, all you had to do was set up one page to reference the other via that serial number. Qualitatively, that's close

to modern hyperlinks. But there's one key difference. Each page could only have a handful of links on it. Now, that's a big problem from the standpoint of a modern-day web surfer. However, Now, that's a big problem from the standpoint of a modern-day web surfer. However, there are some good reasons for the limitation. Part of it is technical. There's a limited amount of space on each rapid selector frame to fit machine-encoded data like links and IDs. You could probably squeeze in enough space for a few links per page, but you'd always have a pretty small upper limit. The limit also comes down to how Bush envisioned Mimex being used. In a website like Wikipedia, for instance,

hyperlinks form a sort of interconnected tree of pages. From one page, you can link to multiple more pages. Then each of those has its own separate set of pages that it links out to. A user can go up or down that tree, moving from branch to branch via links. Part of what makes that possible is the sheer amount of information housed by a website like Wikipedia. As of today, the site has over 50 million pages, and those are put together in an insanely intricate web of links. Although Mimex was meant to house an entire library worth of information, it would

still have been on a totally smaller and different scale than something like the modern web. Besides a limit on the number of links on each page, Mimex was just meant as a much more personal device. Most websites are made by someone else and then shared around, but Mimex was described as a direct and personal machine aid. The owner of Mimex, let us say, is interested in the origin and properties of the bow and arrow. Specifically, he is studying why the short Turkish bow was apparently superior to the English longbow in the skirmishes of the Crusades. He has dozens of possibly pertinent books and articles in his mimics.

First, he runs through an encyclopedia, finds an interesting but sketchy article, leaves it projected. Next, in a history, he finds another pertinent item and ties the two together. Thus, he goes, building a trail of many items. Occasionally, he may insert a comment of his own. End quote. A user wouldn't get a pre-linked directory. Instead, one would build up a personal trail of their research. In that sense, Mimix was, first and foremost, a machine meant to help the time-consuming task of scientific or academic research.

It was never a general-purpose computer. It was a very specific tool. That brings us to the final big feature of Mimex. So I said that the core of the machine was a rapid selector, and so far I've just covered the selection and display part of the rapid selector. But like the earlier MIT version, Mimex was designed to run off copies of selected microfilm slides. Users could copy selections from their library or even make reproductions of their own linked notes to pass off to other Mimex users. One has to imagine that if Mimex was devised just a handful of years later, this would have been replaced with some type of network connection.

So that's a description of Mimex. We have a pretty futuristic device that uses touchscreens, hyperlinks, and shareable data, and it's all designed with 1940s technology in mind. So where can you see one of these amazing machines in action? Well, actually nowhere. That's one of the things that makes Mimex such a strange case. It was put down on paper and the technology to implement it all, at least at some scale, existed, but a copy was never built. Instead, Mimex was presented as more of a thought experiment, and I think that's one of the reasons that in latter years it would become so influential. If it had been built, then many of the ideas behind it would have been bogged down by technical problems, or plagued by mismanagement. By existing in this ethereal space, Mimex was able to serve as a blueprint, and it would be a model to work towards.

I want to close this out with one more quick quote from Vannevar Bush. This one comes from an essay written 10 years after As We May Think. Quote, I believe we shall advance in our mastery of the records we create, rendering them easier to consult by means which now seem strange and bizarre to us. End quote. We're living in a world shaped by the World Wide Web, itself a strange and bizarre technology. In the 1990s, the modern internet would finally take shape, first developing at CERN and eventually spreading around the world like wildfire.

One of the core features that made the internet possible was hyperlinks. The internet is a contagious idea, but it's been that way since long before 1990. I think there's a case to be made that Mimex is where the idea started, or at least that Vannevar Bush was one of the first people to see the full potential of a technology like the internet. From Mimex and Bush, the idea would only spread. And it's an idea that's still evolving, and that we still reap benefits from today. Thanks for listening to Adjunct of Computing. I'll be back in two weeks time with another piece of the story of computing.

And hey, if you like the show, there's now a few ways that you can support it. If you know someone else that would like it, then why not take a minute to share it with them? You can rate and review me 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 access to early episodes, polls for the direction of the show, and other assorted perks. 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.