Vannevar Bush and Ted Nelson Revisited

October 1, 1998


Mary Hopper: I designed this event to address a key goal of the goal of the Media-in-Transition project, which is to encourage a historical perspective of the relationships between digital media and older media. My personal interpretation of this is that perhaps the best path into the future is through the past. Without a historical perspective, we are running blind into the future without a lot of guidance. The past is at least one valuable tool we can use to map what to do more thoughtfully in the future.

My focus today is particularly on hypertext, the corner stone of the World-Wide-Web (WWW) and one of the most "hyped" aspects of the new digital media. This forum is designed to provide a more in-depth picture of what its about. It seems ironic how little discussion surrounds such a key component of our ongoing period of media transformation. Even popular books for novices, such as Internet for Dummies, mention Vannevar Bush and Ted Nelson, but serious books about the Internet and WWW seldom go further than a cursory history. You will almost always find a paragraph on Nelson, and it says he invented the term hypertext. You might find another paragraph on Bush that says he created the Memex. Everything else is almost always left out unless you start digging into serious hypertext literature.

The main thing I am going to do is present some video clips that have very seldom been seen. The clips are from a symposium called As We May Think, A Celebration of Vannevar Bush's 1945 Vision. This symposium provided a unique opportunity to view hypertext from a special historical perspective. The clips from that event show how hypertext, as viewed by Bush and Nelson, evolved from conceptualizations rooted in older media towards the reality of the Berners-Lee's WWW.

Our virtual guest who are in the video clips are Paul Kahn and Ted Nelson. Paul Kahn is a scholar who has written about Bush's Memex, and he was invited to speak at the symposium about his work. Rather than presenting the clips linearly, I have decided to take advantage of hypertext by creating juxtaposed segments from different presentations that are related. People that were actually at the symposium saw one complete presentation by Nelson and one presentation by Kahn, and they never had the opportunity to break it all down and analyze it.




Before I go discuss the backgrounds of Bush and Nelson, I would like to mention the key initial publications that they both wrote about their technologies which are often referenced. Bush's article about Memex first appeared in the Atlantic Monthly (1945), while Nelson's ideas that were the foundation for Xanadu were first printed twenty years later in ACM Proceedings (1965). In addition, they also both then published follow-ups about their technologies many years after their first publications, and Nelson is continuing to write about his concepts today. One interesting point to note is that not only was Nelson definitely influenced by Bush, but in fact, reprinted the entire Memex article.

Both Bush and Nelson were highly connected to traditional media, and these interests played a role in their particular forms of technology they chose to explore. Bush was most famous for overseeing the Manhattan project and establishing the National Science Foundation (NSF). He held many titles, the most well known of which was President Roosevelt's science advisor. He also held the title of President of MIT, which is why the Bush Symposium was at MIT and all his archives are here. Almost always, when Memex is discussed, you get the sense it is a digital tool like a computer, but it wasn't. All of Bush's technologies, including the Memex, were analog devices that were purely mechanical. His earliest invention was the profile tracer, which he patented while a graduate student at Tufts. It is seldom mentioned that he also invented the justifying typewriter. His most significant invention was the differential analyzer, which was used to calculate ballistics tables during W.W.II. Nelson's background was quite different. He focused primarily on the humanities, having earned a B.A. in Philosophy from Swathmore College in 1959 and a Masters in Sociology from Harvard in 1963. His interests began with an orientation towards traditional media like cinema and particularly special effects. However, he was equally focused on the theatrical and theoretical aspects of media.

Broader Concerns

Although these gentlemen were quite different in both their focus and careers, they both shared some broader concerns for what they were doing, and the particular technologies they developed were intended to address these broader concerns. Both of them wanted ways to improve people's ability to store, process and communicate their ideas. That is the commonalty that brings them together. Bush said that the role of the computer in the future would be to "supplement a mans thinking methods." For him to say this was particularly significant when you consider his role in history. To say he founded the NSF and a few other things doesn't really capture the point. He designed both the military industrial complex in this country and the role of information systems within that framework. It is easy to argue that vision underlies what eventually became the Internet. Nelson also has been quite concerned about the nature of the human mind and its relationship to computers. However, his perspective is somewhat different. While Nelson's emphasis has also been on the computer's role in improving the human mind's ability to comprehend complex things, he frames the task of designing media as an art rather than science or engineering.




Let us now turn to looking at what the technologies behind Memex and Xanadu.

Bush's Memex was designed as an analog device, like his other inventions, although it was never actually built. On the other hand, a different machine that was Memex's predecessor, called the microfilm rapid selector, was built in the late 1930s here at MIT. That technology was elaborated over time, until it became quite sophisticated in the hands of the CIA. However, that technology never quite achieved the level of sophistication that Bush described in his article. To give a better sense of what the Memex that Bush proposed would look like to the participants at the Bush Symposium, Kahn provided an animated demonstration of the "Turkish Bow Scenario" that Bush described in his article and a point and click interface to explore the technology behind Memex.

Nelson is credited with coining the term hypertext, but most people think that just means linking like we see on the WWW. In fact, his concepts are more complex than that. In order to give a better sense of how Xanadu was intended to function, Nelson presented his notion of how a computer screen should look, and then demonstrated a piece of software called the Zip Editor to illustrate the key components of his Xanadu concept. In contrast to Bush's "Turkish Bow Scenario", he used the Zip Editor to show how his ideas could be applied to a historical writing example. He spent a considerable amount of time explaining his model of transpublication which would allow the virtual republishing of documents within a new copyright framework. Near the "grand finale" of his presentation, he then summarized the core ideas that are still missing from our current ideas of hypertext.

  • Transparallel - things you look at together whose specific connections are of interest.
  • Transclusion - virtual instance across a boundary with original identity maintained and original content available.
  • Transpointing - allow pointing across window boundaries explicitly.
  • Transvisibility - see from one transcluded instance to another.
  • Transpublishing - virtual republishing by distribution of pointers where materials obtained from the originator or their agent.
  • Transcopyright - permission doctrine.

Issues of Success

In addition to sharing similar goals, both the Bush's Memex and Nelson's Xanadu faced questions regarding their technical feasibility.

Kahn began his exposition regarding the technology of Memex by providing an account of how the diagrams that the public has seen weren't created by Bush. He also explained, not only did the article describe technologies that never did or could exist in the form described, but also that the earlier machine upon which they were based, the rapid film selector, was also later abandoned because of its mechanical shortcomings. Kahn's final point was that Memex was more of a successful exercise in imagination than a technical feat.

Given that Nelson was working with 20 years more of technological development at his disposal, the question of technical feasibility has been a more serious issue that has haunted his career. A complete system like he described has never been built, and there has always been controversy surrounding why. Speculation has attributed the cause to a variety of reasons:

  • The technical concepts were too far fetched to function.
  • The technology simply wasn't advanced enough to support it yet.
  • He wasn't able to communicate it well enough.
  • He didn't have the people skills needed to run the scale of project required.

Nelson directly addressed these issues and attributed the problem to a combination of factors that included both mismatchs in his personality and timing. However, he also explains that he believes that his ideas are implementable on the WWW within a Netscape browser model. In fact, he is still in Japan working on it. While Bush's Memex was clearly more vision than reality, you can see that at least Nelson has the Zip Editor he shows in the clips. Recently, he has also released another piece of software called ZigZag, available on the WWW for free. His attributing the entire problem to himself, while containing some truth, may also be somewhat of a disservice.

Today, there is a great recognition that the complexity required for building large software systems is a incredible problem faced by anyone trying to do it. While he didn't make great progress, he is not the only one. In another presentation at the Bush Symposium in 1995, Tim Berners-Lee said that it was a trivial problem to make the WWW interactive. That was almost five years ago. Anybody that develops for the WWW today will still tell you that it is not terribly easy to edit the WWW or to get other people to interact with each other in the same workspace. That is a really major theme right now. There are grounds to suggest that, while the WWW has been a wild success in some ways, some of Berners-Lee's most valuable larger goals were also never reached.



Paradigms Missed...

It is easily argued that Bush's vision of Memex has been realized in spirit, while much of Nelson's vision remains unattained by him or others. Nelson eloquently described the limitations of some of our favorite interfaces. In a wonderful series of clips, he attacks the Macintosh for being a paper simulator with inadequate metaphors like cut and paste or the garbage can. Nelson continues to remind us that we still need more elegant interfaces to help us to better express the multidimensional nature of our ideas.

Now, as Nelson suggests, let us consider the possibility that there are many things that we could potentially do electronically which we have never achieved. Hypertext, like computing in general, has a long history of valuable features that have been implemented in demonstration systems, but have not made their way into well known commercial systems for the general public. HyperCard was definitely considered a bastardization of the hypertext concept when it became popular, because there were much more sophisticated hypertext systems available long before it was released. Many of the features from early systems, as well as many newer concepts, are still not available in any widely available systems -- including the WWW. We should wonder whether there is a large percentage of the possibilities of hypertext, or computers in general, which are not available to the average person yet.

Paradigms Revisited

Consider that there may be vast potentials in electronic media that we still haven't implemented, and we may be in danger of missing some of the greatest potential in electronic media because we mistake some situational limitations in technical development or project management for impossiblity. This is the grounds upon which I suggest that the past is of more than just historical interest. The past, and the pioneers who shaped the past, may hold many suggestions as to paths it might be valuable to explore in the future and advice on the most productive ways to go down those paths.




Mark Bernstein: First, let me begin by mentioning that I wasn't around for the early history of hypertext. I started building systems in 1982, almost twenty years after Nelson. I did predate HyperCard, but I came in later than Nelson and his early Xanadu.

I have opinions based on insubstantial things about Bush. If Bush's essay had been forgotten, as it effectively was when Nelson reprinted it, would anything have been very different? People will say things like, "Bush created Memex." But, of course, he didn't, because it never got created. Its not even a failure of management or implementation. As far as I've been able to discover there was never a serious effort to create it. Memex doesn't really anticipate Xanadu in a very meaningful way. Their interests are very distinct and different.

For many years, there was a tendency to believe that Nelson was a starry eyed leftist idealist whose ideas were simply impractical. This is now very hard to recover -- especially if you enter the world post WWW. To countermand this, Nelson adopted a very business oriented rhetoric in the early 70s. This is why he fell into branding every term. Notice that every single concept had to be "trans" something. This led to Nelson getting into a content free, but no win bind in his speaking and rhetoric in the 80s. On the one hand, he is too business like, because he keeps branding everything. On the other hand, he is not business like enough. So he could never raise enough money to build it all.

As late as Hypertext 91, we had a stringently referred paper in a terribly competitive conference. It was a paper analyzing the transcopyright model, and it said, "in conclusion, most uncertain are the adequacy and financial incentives for authors to put their most valuable copyrighted works into the Xanadu system." This seemed like a completely reasonable assertion as late as 1991. Of course, by 1993, people were lining up in droves to pay people to put their copyrighted works on the WWW, which for all practical purposes was the Xanadu vision.

There is also the political issue that is important to remember, but wasn't discussed much. The founding document by Ted Nelson that everyone has in mind, especially those who came onto the scene after 1987, is a book called Computer Lib (87). One of the two front covers in that book is a wood cut with a clenched fist surrounded by the slogan "computers for the people, you can and must understand computers now."

I think this is the decade in which computer liberation actually becomes significant as a political force. They have stopped being things that we treat with reverence like they were a decade earlier when we were told we could hand them offerings through the window.




Question: I'm beginning to see that what Nelson was getting at was the incredible problem of trying to build software to accommodate human thought. What is the next step?

Bernstein: The next step is always going on in lots of places. Look at the systems that many people are working on.

Hopper: What are some of the biggest things that have been left out of current systems.

Bernstein: Everyone knows that the worst thing about the WWW is latency. Even The New York Times knows it is a problem. How many hours are lost? Today, we seem to be falling into using the WWW as a synonym for hypertext. That's probably a mistake, a blip in our consciousness. A lot of hypertext happens off the WWW, and it will continue to do so. We are also going to see some interesting hybrids. For example, I expect that, in a year or two, the usual way to buy a Eastgate hypertext will be through the wire. It will be buying a chunk of stuff that will show up on your harddisk and stay there. There are a lot of things about the browser that are dictated by latency or just wrong.

The biggest blunder is that links are underlined text. This was only one of a whole host of proposals of representing what links should look like. Some people used typographic indicators, colors or symbols. Almost anything would be better than colored and underlined, because colored and underlined reads like emphatic. There are lots of things you want to do with typographic emphasis that have nothing to do with saying there is a link here. A classic problem in footnote writing is that they always call undo attention to themselves. The footnote that says "I stole this from somewhere" is indistinguishable from "the foregoing passage has been shown to be completely wrong, and if you build a bridge based on it, it will fall down."

Copyright is also a big problem. What we have now is an improvised cronieism from the seventeenth century that has worked well over the years, but that doesn't mean it was handed down on stone tablets, and we can't revise it. I have had good fortune, but it can be a insurmountable problem. I know lots of people have things they badly want to do, where they are facing viable copyrights that are being defended by people who badly want to see the work not happen.

Alexander Chislenko: If you want to improve on existing systems, as in other areas of engineering, there are two ways to do it. One is to figure out what the good system should be and build it from scratch. The other is to take advantage of the existing conditions and build on them. This would mean taking into account the huge existing WWW and adding elements that would be something more to our liking. I see some elements of this going on, like semantic markup in the form of lets say XML, and I wonder if you think the existing WWW can be mended.

Bernstein: Remember, I am not Ted Nelson. I think the WWW is great! In fact, so does he. He has actually written an article that I am hoping to publish called "Two Cheers for the Web". Things that would be nice on the WWW would be browsers that would work if they are reconfigured. HTML is a great markup language, but now we know all the things that it messes up, such as the confusion between visual and content markup.

The important distinction is not all the things that HTML did wrong, but that there was a perfectly good markup language before it called SGML. Lots of people thought it was going to be the future, but it was an engineering disaster. This was because, when they built the standard, they apparently had representatives from every international agency you could ask, but, apparently, zero computer scientists. So, they wrote a language that wasn't parsable, and that's no fun. On the other hand, HTML can be parsed by a second year undergraduate on a bad day, and some of that code is still in Netscape. That's the big answer. Lots of the things that people want to do could be added with tags, tag extensions or fancy servers. Its easy to sit and say Microsoft or Netscape should do this, but remember this all just happened yesterday, and it happened over night, and it can change.



hypertext in historical context    abstract    speakers