Replacing the Printed Word: A Complete Literary System

Theodore H. Nelson

Developments in computer technology now make it possible to store stupendous amounts of written material economically. This leads to radical visions of what may come to be. However, if paper is to be supplanted, it must be by something rather like the paper systems we now have, even if in some abstracted and unfamiliar form. What, then, have we? And how to abstract and extend it? Our model is "literature"-defined in a somewhat unusual way. Under guiding ideas which are not technical but literary, we are implementing a system for storage and retrieval of linked and windowing text. The "document", our fundamental unit, may have windows to any other documents. The evolving corpus is continually expandable without fundamental change. New links and windows may continually add new access pathways to older material. Fast proprietary algorithms render the extreme data fragmentation tolerable in the planned back-end service facility.

THE POSSIBILITIES AND THE PRECONCEPTIONS

Vast text storage, and communication techniques for reaching into it, are new technological developments to be coped with--which very few people were expecting. So-called "word processing" systems have already greatly changed the handling of the written word in the office environment; now the question is how to extend this approach. It now is feasible to replace large-scale systems of in-house publishing with computer storage (Lancaster, 1978); and it will soon be practical to do the same for publishing to the open public. Indeed, true electronic publishing is feasible now-- that is, public-access document systems with digital storage and demand service to the open public. The problem is not electronic. It is not "software", meaning procedural obstacles to implementation. The problem is conceptual. If such systems are to be promulgated to a wider public-- no longer just in-house-- they must be clear and simple to use, yet offer powerful new features. They may not merely be a clumsy imitation of paper systems. This is a design problem. It is a problem of creating a conceptual structure, an organizing system of ideas and methods and patterns of work. There is no universal or obvious approach to this problem, though numerous candidate approaches exist. Various parochial disciplines and ideologies within the computer field, and other fields, have styles of thinking that seem to speak to this problem. How clearly they speak to it is a matter for careful thought. In word processing, though screen methods for actual edit have become streamlined, all systems appear to bog down at the borders of a document. The conventions of computer storage are little improved, and so secretaries must deal with all the usual file conventions, problems of file naming (and keeping track of file names), backup, and keeping backup copies safe and segregated, and so on. Few linkage facilities exist either within or between documents.

A number of on-line communities exist, particularly around such time-sharing systems as the ARPAnet. But they, too, suffer from the conventional problems of file naming and backup, in addition to the greater complexity of using the systems at all. Curiously, the users of such systems still must in most cases give their documents short names, keep paper lists of these names, and juggle backup copies in the same way that secretaries do. Systems for "electronic mail", though much publicized, are still largely in the telegraphic tradition. The "office of the future", we are told, is quite near. There seems to be little agreement on what it will be, however. Some say it will have optical scanning for input and paper for output. Some say it will have an "information manager" who will try to keep it all working and keep the system brought up to new revised specification. All the time. Some say it will not even happen, whatever it is. Some researchers, impressed by the work of Douglas C. Engelbart and his "NLS" system at Stanford Research Institute, have proceeded on the assumption that good tools for the intellect can be built with good text systems. The NLS system is essentially a community of shared files, with facilities for rapid search and linkage. As implemented it requires extensive training; but Engelbart's ideas have been widely influential, and would seem to lead toward simpler systems for wider user groups(2.3.4) A recent development has been the teleconferencing system, most conspicuously developed by Murray Turoff at the New Jersey Institute of Technology. As most often seen, the teleconferencing system is a setup for accumulating messages sequentially from different participants. These messages, always added at the end of the accumulating scroll of text, tend to take on a character not unlike informal conversation. In some versions, however, they have become complicated by tricky sets of rules and an elaborate supervisory function for those people designated as conference leaders. The field of "information retrieval" appears to have stabilized into a certain basic form. This is usually the storage of a number of items of text which can be scanned on the basis of a formalized user query. (A commercially successful example is the New York Times data bank.) Usually what are stored are article summaries, but this approach could in principle be extended to fulltext as well. However, the training to use the search and query methods of such systems tends to be long and arduous, the cost is high, and users express disgruntlement about the unpredictability of results. Advocates of artifical intelligence have repeatedly informed us that "everything we want" will be forthcoming in an unspecified manner at an unspecified date in the very near future. However,we are assured that it will involve interactive dialogue with some kind of intelligence whose internal workings and system of thought need not concern us. This Softbeing will understand us perfectly, foresee our needs, wish to keep us completely satisfied, and have nothing to hide. In summary, I would say that the situation is one of total confusion and odd preoccupations. I see no obvious, let alone conceivable, way that all these concerns and obsessions can be comprised into a single outlook, let alone a common system. (Nevertheless, there are respected researchers, such as J.C.R. Licklider, who seem to think they can (5).) Perhaps the apparent complication and mutual intractability of these different approaches is related to something else: they all have rather little relation to our present uses of paper.

INTERACTIVE SYSTEMS AND THE DESIGN OF VIRTUALITY

Our approach to a computer design we call "the design of virtuality." By virtuality we mean the seeming of an object or system, its conceptual structure, its atmospherics and its feel. Every object has a virtuality, a seeming. Natural objects are more or less what they seem to be; man-made objects are not. The virtuality of a house, or an automobile, is what the designer made it--the structure and qualities that were chosen, and the techniques by which they were realized. The closest analogy to the design of interactive computer systems, I think, is the making of movies. What counts is effects, not techniques. We are not concerned with just how a certain effect is to be achieved, so much as with what effect is wanted. An effect is something intended to take place in the mind. Suppose the movie effect desired is a sense of a monster approaching. This can be done by showing a man in a lizard suit--yaargh--or animated puppets, or by showing the fright of a person who sees the monster. In other words, a variety of techniques may be selected toward a common effect. The design of an interactive computer environment, similarly, should not be based on particular hardware, or a particular display device, or a programming technique. It should be based on the intended effect. in the mind and heart of the viewer. ("Heart" here is added because we are too seldom mindful of the emotional component in a user's reaction.) Another way of saying this is that the "systems analysis" for an interactive system should deal with the mental space of the user's experience. The process is a cycle: study, and design. First we must study the approximate structure of whatever we are designing, and roughly what it is about. Then we design, that is, look to see how the computer's capacities may be made to assume a similar conceptual shape.

There is one other key constraint in system design: conceptual simplicity. If any but highly-trained people are to use a system, it must be extremely simple. It must be simpler by far than anything computer people are accustomed to designing--a factor of ten, let us say, simpler than what a computer hacker considers "simple." Popular lore in the computer field holds that simple systems are not "powerful" -- where powerful seems to mean "allowing concise macro-language programming." (This is evidently the view of those who consider TECO a powerful text editor, or, indeed, a text editor.) We believe that true power, meaning easy and focused control by the user on what he means to do, is not merely compatible with simplicity, it requires it.

This has been a preface to understanding our design. Our approach to electronic text has been to look for the hidden nature of writing as a whole, and the way it is used, to find a paradigm for this design. The virtuality we have designed reflects this endeavor. (In point of historical fact, this philosophy of virtuality and simplicity arose in parallel with the development of the design, but we think we now know how to design any other system on the same basis.)