Ted Nelson

Ted Nelson is a strange man. He has a tendency to self-aggrandization. He’s claimed to be the inventor of things he only predicted, with no hand in their construction. He is unlike most revered figures of computing history. We idolize people like Donald Knuth, who contributed to the field of algorithms and created programs like TeX; Dennis Ritchie, who helped to create Unix and the C language; Grace Hopper, for her early work on compilers and languages. All these people developed knowledge and programs we still use today. Nelson doesn’t have such a thing to his name; his tangible contribution to computing is less significant than say, the writers of the MD5 standard, or the average volunteer kernel developer. It’s safe to say, he is a vastly overrated historical figure in the development of the computer or the web; he is more or less a hanger-on to figures like Douglas Engelbart and John C. Lilly. He is still alive today, and he still projects himself to the world on YouTube, as a bitter old man desperate for recognition.

I make an unflattering picture of Nelson; I am tired of those who call him a ‘visionary’ when he did nothing. He put this word into the mouth of others by tricking others into believing him important. Yet for all this, I will never deny Ted the rightful credit to an extraordinary concept that I regret never came to be; that is, Project Xanadu.

If you are not aware, Project Xanadu is a concept of hypertext which Nelson created, which is far more advanced and interesting than what we ended up using, which is HTML. I’m not aware of any proper Xanadu system that is not purely demonstrative; it is most infamous as a story of great failure, which is unfortunate because this overshadows what a brilliant concept it is.

What Xanadu is

Certain ideas of Project Xanadu, namely interconnected media sharing, were realized by the world-wide web, though in their time were revolutionary. Today though, there remains at least two ideas of Xanadu that have not been done, which are it’s most interesting. The first is that a text can have a contextual link to another text, which facilitates the client displaying the referenced text in parallel, with the contextually linked sections of text shown graphically by a drawn line and text highlighting. There is also the concept of transclusion which I like most; a type of link that doesn’t just point to content, but will retrieve it and displays it in-place in the document.

Suppose you are reading a book on theology; lets call it ‘Companion to the Bible’ by John Doe. You are reading it as a Xanadu document (Xanadoc). Companion makes a reference to a line from Psalms; instead of embedded it in the Companion file, it finds this line from a different document of that Bible, brings up that portion of Psalms besides — in parallel to — the section of Companion referencing it; may also highlight the text and draw a line between it. In fact, within that paragraph of Companion, it may have many such contextual references, so the lines help visualize the textual connections.

This may not seem particularly impressive, important, or useful. I believe that if this was really tried, then it would become clear to us how intuitive and useful this actually is. If you think about the implied architecture of this system, you can think of even more possibilities.

Applications of Project Xanadu

Syntopical Reading

In “How To Read A Book”, Mortimer J. Alder proposes four types of reading, each of which transcends the prior form of reading and has stronger intellectual potential. The fourth and greatest type of reading is ‘syntopical reading’, which is the reading of many books concurrently — in parallel — and the reader makes links between them to develop insight in the subject. This is difficult, and part of that is wrangling with the medium of linear books, because we unfortunately retained the trappings of physically limited books in their digital form. Project Xanadu is an elevated medium that could be useful for syntopical reading.

Mathematics Textbook

Imagine a mathematics textbook with theorems and lemmas that are within or without the book. Instead of flipping between pages or hunting down the referenced text, you could instantly display them besides the proof you are reading, whether this is in your current book or another one.

History Textbook

Imagine a history textbook, all of it’s sources, footnotes, endnotes, bibliographies. If you wish to traverse the sources, though it’s easier in the digital age, it is still not as effective as possible. A Xanadu system could allow the traversal of these sources instantly! Note that a Xanadoc doesn’t have to be text; they can be visual and audio media as well. For instance, museums could provide media of artifacts for the Xanadu network.

Technical Documentation

Consider a new standard, beginning with Version 1.0 as a complete document. Version 1.1 comes along; rather than being a whole new document in itself, the relevant sections of 1.0 can be transcluded, and new parts interposed within the document. Decades pass, we come to version 6.0 of the standard. The structure of such a Xanadoc file would literally encode the evolution of the standard, showing what exactly changed when. Evaluating what work is required to upgrade a system from v4.3 to v6.0 is only a matter of viewing the file difference between the two Xanadocs! Then, you may write your own internal documentation for systems as Xanadocs, which make parallel or transcluded link to the standards documentation, allowing the reader to immediately understand the relevance of something.

and many more!

I hardly need to describe how useful this system would be if it superseded the PDF as a form for scientific journals. And note that anyone could write their own Xanadoc; imagine how useful this would be for note-taking, knowledge management, etc. not just for a student, but also for writers of many sorts.

Creating a Xanadu implementation today

Some of the challenges I see in a Xanadu implementation today are such:

• There is no compatible content to begin with, thus no motivation for users to use it. This challenge will be ever present in the face of non-universal adoption.
• Overcoming the ’link rot’ that plagues modern systems. This is less of a challenge for books with an ISBN or standards created by specific organizations. What of a Xanadoc without any such identification?
• Security is a massive concern. Xanadu implies authentication for delivery of copyrighted material, and possibly financial transactions for permission to transclude/reference material.

I’ve thought a bit about how Xanadu could be implemented. I imagine there would be clients, and at least two kinds of servers.

• Publishers: publishing companies, museums, galleries, organisations; provide Xanadocs directly.
• Distributors: universities, libraries; can request and cache documents on behalf of clients.

Clients could request documents from publishers, but may have to pay some fee. If you work for a university, you may have access to all sorts of material through your university library. This is all very fanciful speculation on my part.

The Disappointment of Hypertext

The chance of Xanadu taking off in 2025 or beyond is close to zero. It is a reminder that we haven’t realized everything computers can do. Xanadu is a fundamentally difference approach to hypertext, not unlike how Lisp and APL are fundamentally different approaches to programming; they were all created extraordinarily early in computing history, and are beloved even today for their bold concepts and underappreciated potential. But at least Lisp had it’s moments through history; Xanadu never got a chance to really show itself.

When we created digital documents, and also webpages, there was so much we could have done with them. To borrow from Baudrillard, these ‘documents’ are not necessarily with any origin, they need no reference point. Such an item of cyberspace could have been created with whatever feature and form we wanted to give them, hence we could have had ‘hypertext’; that is, text that goes beyond the limitations of text we’ve experienced so far. HTML is barely a hypertext; all it does is make pedantic, graphical commands, and does not give new form to the document. The only thing ‘hypertext’ about it, is the inherent property of infinite reproduction of digital data; how lazy! by itself, HTML does not make any new thing possible. The PDF, DJVU, and other document formats, also retain the old trappings of documents, specifying graphical rendering of a linear, non-interconnected digital ‘object’. All the digital document came to be, is the perfect rendering of the classical document with digital data; what a blunder! Now, we have ‘web apps’ that has turned the hypertext standard into something of a graphics assembly language for a virtual machine we commonly refer to as a “web browser”. All of this to say; the standard form of ‘hypertext’ HTML rejects new possibilities, and hardly lets us do things not otherwise possible. We preferred things as they were; hence, Microsoft Office interface is covered in familiar icons of paper, scissors, pencils, and such; we eschew real conceptual progress and prefer familiarity. It seems like we just don’t want hypertext. Maybe this is the cosmic reason why Xanadu could never have come to be, no matter what we did.