Theodor Holm Nelson, born 1937, obtained his BA in philosophy from Swarthmore College. In 1960, he was a masters student in sociology at Harvard. Shortly after enroling in a computer course for the humanities, he was struck by a vision of what could be. For his term project, he attempted to devise a text-handling system which would allow writers to revise, compare, and undo their work easily. Considering that he was writing in Assembler language on a mainframe, in the days before "word processing" had been invented, it was not surprising that his attempt fell short of completion. Five years later, he gave his first paper at the annual conference of the Association of Computing Machinery (ACM). It was around this time that he coined the term "hypertext."
Since that date, Nelson has been pursuing his dream, a software framework he named Xanadu, after Coleridge's "Kubla Khan" (he came up with the name while working for a publisher). This he describes at length in Literary Machines, calling it a "magic place of literary memory" (1/30).
The Xanadu software is as mythic as the place after which it was named. In Dream Machines, published in 1974, Nelson announced that it would be ready for release by 1976 (56). In the 1987 edition of Literary Machines, the due date was 1988 (0/5). The development of Xanadu was given a large boost in early 1988 when Autodesk (the company which made their fortune from AutoCAD) bought the Xanadu Operating Company. Code for a prototype of part of the system was made public later that year. In an article published in Byte in January 1988, Nelson expected to be fully completed by 1991 (299). Then, nothing. Autodesk has since relinquished interest in Xanadu.
Nelson's conception of hypertext is a rich one. Dream Machines describes hypergrams (branching pictures), hypermaps (with transparent overlays), and branching movies, such as the film at the Czechoslovakian Pavilion at Expo '67 (44). The modular layout of this book attempts to impart the interconnectedness of knowledge which hypertext can convey. Its large format, hand-drawn illustrations, and irreverent tone were inspired by Stewart Brand's Whole Earth Review. Flip the book over, and you'll find a second polemic--Computer Lib. The book sold a total of 50,000 copies.
In Dream Machines, Nelson provides three categories of hypertext (45). The first, basic or chunk hypertext, supports what we have been calling reference and note links. The second, stretchtext, is a full implementation of expansion links. The third, collateral, stems from his work in 1971 with the Parallel Textface, which provides a view of two documents on one screen, with full support for versioning. Nelson also distinguishes between "fresh" or original hyperbooks on one topic, "anthological" hyperbooks linking different works, and "grand" systems:
[These consist] of "everything" written about the subject, or vaguely relevant to it, tied together by editors (and NOT by "programmers," dammit), in which you may read in all the directions you wish to pursue. There can be alternate pathways for people who think different ways. (Dream Machines 45)
This vision obviously owes a lot to Vannevar Bush. Indeed, Nelson reprints the entire text of "As We May Think" as a chapter in Literary Machines.
A penchant for coined words and vague generalities obscures the substance behind Nelson's claims. This has resulted in some confusion over what, exactly, Xanadu is. Most references to Nelson excerpt one short quote, and then move on. He has not failed to notice this reaction, writing:
The project is well known, but not well understood. Its greatest aspiration, a universal instantaneous hypertext publishing network, has not been generally understood at the technical level and has created various false impressions. One publication, for example, referred to it as "a database-to-be the size of the world"--a very muddled description. ("Managing Immense Storage" 225)
Nelson then goes on to talk about xanalogical storage, humbers, the docuverse, and tumbler arithmetic, not exactly making it easy for the general readership to understand his ideas.
Buried beneath the terminology are some important concepts. Humbers are HUMongous numBERS, arbitrarily large and based on a forking number system. Humbers were devised so that an infinite number of unique IDs may be generated for labelling stored text.
Xanalogical storage, which Nelson later termed transclusion, describes the ability to make a virtual copy of part of one document, for inclusion in a second document. The original remains untouched, in place. The link between the source and target documents is maintained throughout further operations. Nelson sees this as a means to get around several problems, notably copyright and transcription errors:
[T]he customer is buying the fragments from the original author whenever those quotations are read. So, nothing is misquoted, nothing is out of context, credit is apportioned correctly, and royalties are apportioned correctly. ("On the Xanadu Project" 298)
On the face of it, this analysis is highly suspect. A small fragment of an original work could still be used out of context. The byte-by-byte royalty method rewards quantity over quality. Nelson's understanding of copyright, as it currently exists, is flawed. And his ideal of a McDonald's-like information service is less than palatable.
Despite Nelson's unwavering optimism, Xanadu has failed to materialize. Nonetheless, its intellectual presence has exerted an enormous force on the evolution of hypertext systems. Few researchers would deny the influence of his ideas.
[1995 update: Recently, there have been rumours of Xanadu Light, a hypertext system incorporating at least some of Nelson's concepts. Whether this sees the "light" of day or not is irrelevant; the World Wide Web embodies many of these ideas, and is available here and now.]