THE LITERARY PARADIGM A piece of writing--say, a sheet of typed paper on the table-- looks alone and independent. This is quite misleading. Solitary it may be, but it is probably also part of a literature. By "a literature" we do not mean anything to do with belles-lettres or leather-bound books. We mean it in the same broad sense of "the scientific literature", or that graduate-school question, "Have you looked at the literature?" A literature is a system of interconnected writings. We do not offer this as our definition, but as a discovered fact. And almost all writing is part of a literature. The way people write is based in large part on these interconnections. A person reads an article. He says to himself, "Where have I seen something like that before? Oh yes--," and the previous connection is brought mentally into play. Consider how it works in science. A genetic theorist, say, reads current writings in the journals. These refer back, explicitly, to other writings; if he chooses to question the sources, or review their meaning, he is following links as he gets the book and refers to it. He may correspond with colleagues, mentioning in his letters what he has read, and receiving replies suggesting that he read other things. (Again, the letters are implicitly connected to these other writings by implicit links.) Seeking to refresh his ideas, he goes back to Darwin and also derives inspiration from other things he reads--the Bible, science fiction. These are linked to his work in his mind. In his own writing he quotes and cites the things he has read. (Again, explicit links are being made.) Other readers, taking interest in his sources, read them (following the links). In our Western cultural tradition, writings in principle remain continuously available--both as recently quoted, and in their original inviolable incarnations--in a great procession. So far we have stressed some of the processes of referral and linkage. But also of great importance are controversy and disagreement and reevaluation. Everyone argues over the interpretation of former writing, even our geneticists. One author will cite (or link to) a passage in Darwin to prove Darwin thought one thing, another will find another passage to try to prove he thought another. And views of a field, and the way a field's own past is viewed within it, change. A formerly forgotten researcher may come to light (like Mendel), or a highly respected researcher may be discredited (Cyril Burt). And so it goes, on and on. The past is continually changing--or at least seems to be, as we view it.

There is no predicting the use future people will make of what is written. Any summary, any particular view, is exactly that: the perspective of a particular individual (or school of thought) at a particular time. We cannot know how things will be seen in the future. We can assume there will never be a final and definitive view of anything. And yet this system functions.

In other words, even though in every field there is an ever-changing flux of emphasis and perspective and distortion, and an ever-changing fashion in content and approach, the ongoing mechanism of written and published text furnishes a flexible vehicle for this change, continually adapting. Linkage structure between documents forms a flux of invisible threads and rubber bands that hold the thought together. Linkage structure and its ramification are surprisingly similar in the world of business. A business letter will say, "In reply to your letter of the 13th..." or a business form, another key communication, may say in effect, "In response to your order of the 24th of last month, we can supply only half of what you have asked for, but can fill the rest of the order with such-and-such item from our catalog." All of these citations may be thought of as cross-linkages among documents. The point is clear, whether in science or business or belles lettres. Within bodies of writing, everywhere, there are linkages we tend not to see. The individual document, at hand, is what we deal with; we do not see the total linked collection of them all at once. But they are there, the documents not present as well as those that are, and the grand cat's-cradle among them all. From this fundamental insight, we of Project Xanadu have endeavored to create a system for text editing and retrieval that will receive, and handle, and present documents with links between them. We believe there is something very right about the existing system of literature; indeed we suspect that there are things about it that we don't even know, as with Nature. And so we have tried to mirror, and replicate, and extend existing literary structure as we have here described it. But as we said earlier, the design of virtuality is a dialectic. First we must study what there is; then we must venture a design in response to what is. And that design is not dictated by what we have seen, any more than the design of a house is dictated by a hillside. There are only hints.

THE SYSTEM DESIGN By the "design" of the system we mean its conceptual architecture: the basic ideas of it. But with many systems the general system of concepts is filtered through innumerable complications and accidental features. In this system the basic ideas are the only ideas. Our approach, then is to consider all documents as interconnected, or potentially interconnected. Not just documents, but linkages as well, are the fundamental elements of the system. (Links are actually parts of documents, as will be seen later.) It is the different types of links that give both power and complication to the system. Our system, the Xanadu Hypertext System (tm), is intended to store a body of writings as an interconnected whole, with linkages, and to provide instantaneous access to any writings within that body. For this system we have formulated a basic linkage structure that we think meets all the needs of literature as we understand it, including business literature. Beyond this general idea there is little further to expound, except the link structure and the document conventions, which are all the user really has to learn. Since the network is essentially invisible, all we have to discuss are these links and links and documents.