From A Console To A Network

So far we have more or less assumed a single-processor version of this system, one which easily treats all documents and their versions as an interconnected whole because they are stored in the same place. But given today's network technologies, this is not a real restriction.

The system should be able to grow without size limit, containing in the body of available writings whatever anyone has stored from any place on the network. A user at any location may store what he or she wishes; links may be created by anyone, from anywhere, to bring a document (or part of one) to the inquiring user.

All of storage near and far becomes a united whole-- what is now called a "distributed data base." Actual locations are essentially invisible to the user; or, in that traditional phrase, "You don't care where it's stored." The documents and their links unite into what is essentially a swirling complex of equi-accessible unity.

Multi-Person Use

For the time being we will ignore the problem of privacy, and assume that all users are freely sharing their work.

Anything stored by one user on the system may be quoted-- adopted into a document-- by another person writing on the system. No copy is made of the quoted material; rather, a quote-link symbol (or its essential equivalent) is placed in the quoting document. This quotation does not affect the integrity or uniqueness of the original document, since no copy is made. Nor does it affect the ownership: in our planned service, a standard proportional copyright fee is paid automatically by the user every time a fragment is summoned.

The use of special links dramatically simplifies a host of problems.

No copying operations are required among the documents throughout the system, and thus the problems of distributed update, so familiar throughout the computer world are obviated. Since quoted material only has to reside in its place of origin, and not in the other documents that quote it, other files that quote it are automatically "updated" when its owner changes it.

Philosophy of the Perplex

Often the truth about a subject is difficult or impossible to find, though a great deal of information about it is on hand. Frederick C. Crews has implicitly proposed the term "perplex" for such a body of information, in his masterful academic satire, *The Pooh Perplex*(11).

Intuitively, we ought to be able to use computers to help us sort out and order the complexities of what is written, so that our grasp of it becomes firmer and clearer. I have proposed the term "thinkertoy" for such a facility (10); more recently such terms as "decision support system" have appeared in the literature. But what has been less clear is the nature of what such systems should be like.

Our system facilitates multiple interpretations of the same material. Whatever is stored may be seen as a compound object, either organized in different versions or viewed through other documents. Each may independently represent a different point of view. Thus users may highlight different interpretations of the same material, by quoting or linking in from different documents. Thus it seems to be a genuine thinkertoy for strong and subtle intercomparisons.

We spoke earlier of the unending change of ideas, the way in which a given field is constantly subject to reinterpretation. It is a tradition of western thought that such reinterpretation is always possible, always going on. But how can a literature that has been described in one way be reinterpreted in another without total rewriting?

Within our system, the user may make marginal notes and new documents that ease his task in totally reinterpreting (or "newly interpreting") the material before him. He has only to make his own summary of another piece of writing writing-- and indicate the pathway by a correlink. A reader may then summon the corresponding part at any time he wishes to confirm the interpretation.

The improved visualization and control of alternative theories and viewpoints, of disparate and corresponding ideas, should give a person a broader grasp of everything he reads and thinks about. By providing such tools ready to hand, we think we are contributing to a new way of seeing.

A Radical Implementation Sequence

The system as it is currently being implemented is based upon this structure.

Since conventional operating systems work with conventional files, new approaches had to be found.

Now, the normal way to create a new program in the computer field is to it on an existing computer setup. A computer setup consists of a computer and its main control program, or operating system, which handles all file access and update. However, since our approach required a radical redefinition of file storage and operations upon what the files contain, the focus of the enterprise has been on the redesign and re-implementation of what most computer people have thought was completely settled, the file system. This in turn requires the creation of a whole new operating system, since the operating systems presently available rest upon a conventional view and implementation of file structures. These have been non-trivial obstacles.

The current version is now in code in a suitable systems language, and we expect to demonstrate the major functions of the system in a reasonable time.

Publication Through The System On The Postulated Network

The system's design is a unified whole, but we may think of it as the basic conceptual structure, plus a technical structure which makes it possible, and a contractual structure which makes it possible for people to use it confidently. These aspects taken together make a unified design. Because the conceptual structure required very fast lookup within a tightly organized but largely linked system, we had to develop a particular technical structure; and because the conceptual structure expects participants to behave in certain ways, these are embraced in the contract offered to users. These provisions are necessary for the orderly and confident use of published material by many people.

Beyond its use as a private network, we intend that this system be usable as a publication system. Thus a carefully designed system of publication, much like that of paper, has been worked out.

Any user may store anything on the system. Unless specified otherwise, it is a private document. If the user chooses to publish it, however, he may do so with relative ease, making it available to anyone throughout the network. It is then a published document.

Because publication is an important act, both for authors and readers, we make publication a solemn event, to be undertaken cautiously. The author signs an "I hereby publish" form, after which not only is the document universally available, but its author may not withdraw it except by lengthy due process. (He may readily publish a superceeding document, but the former version remains on the network.)

An author who wishes to render his work universally available, but wishes also to retain the right to withdraw it at any time, has a simple means for doing so. He simply designates his document as a private document with unrestricted distribution. Anyone may have access to it or use it, but the owner is free to withdraw it or change it irrecallably at any time.

Any user may read, or otherwise employ, any published document on the system, or any private document to which he has legitimate access. He can make any kind of links to it from his own documents, private or not.

Accessibility and free linking make a two-sided coin. On the one hand, each user is free to link to anything privately or publically. By the same token, each author of a published work is relinquishing the right to control links into that work. The relinquishment is part of the publishing contract.

The user may employ any terminal, graphical or printing. Viewing methods and manipulations are up to the terminal designer. No restraint is contemplated as to what use may be made of the materials found on the system, since no restraint is possible.

We will recommend certain programs for use on the user's terminal or personal computer, but these may be created by any party. Approved terminal programs may be offered certain trademark advantages, but no terminal behavior can or will be forbidden.

Noncoercive In Use

We see this system as offering remarkable power to all users with the greatest possible freedom of use.

The System's Future And Sweep

We have created this system intending to offer a viable alternative to all forms of reading, writing, archiving and study now handled by methods of paper. Through the system it is possible to mimic, perhaps viably, many aspects of the great society of paper: books, magazines, private notes, copyright, royalty, archiving, and roles for author, publisher, and critic.

We want it available to everyone at $2.00 an hour, and to assure freedom of speech on the system, the integrity of copyright, and other high-minded objectives with respect to its broad future use.

Obviously this is a somewhat ambitious plan, and we cannot judge its viability ourselves. But even given more moderate goals, it seems to be a versatile structure for other purposes. For instance, it would seem to be a good public-access memory service, offering a backend with much greater flexibility than standard storage. Objects can be stored "seamlessly," and the need have no concern for their size, naming of alternative versions, linkage and the like. Thus it would seem also to be a favorable storage structure for naive user systems of all kinds, by its removal of various levels of complication.

It should go without saying that the system may be used for all other forms of linked data base, including animated graphics, which we hope will be an important component of future educational and leisure systems.

Such an approach offers to standardize, not languages or processors or algorithms, but the storage form of linkable and complex materials, and terminal interfaces for their exploration. This would seem also to be a worthy goal.

It may be noticed that the system adopts readily for purposes of "electronic mail," using the null adaptation.

Conclusion

Text is the self-portrait of human thought; more precisely, it is the ordered presentation of the results of that thought. Specific textual conventions have evolved in different aspects of human endeavor, but to study any of the by itself is misleading, like studying only one part of the body or only one sex of a species.

The computer field has gradually become aware of text problems, but most computer people see them as independent areas, like "word processing" and "electronic mail" and "information retrieval."

And so it is that computer people have for the most part never looked at the whole picture. Conventional system designers have approached small subsets of the grand text problem, and the resulting designs have tended to be complicated and cumbersome. Often they require, not only a tangle of specialized and user languages, but new social roles for supervisors and service personnel, since ordinary people cannot be taught their use. System complexity appears to rise in proportion to system size, or worse, by some power of system size. But a little thought shows at once that this cannot be permitted. At that rate nothing is going to work.

The view I am suggesting is that the problem is unique, singular and enormous, and the solution can therefore only be unique, singular and enormously simple. I think there are not many text problems, but one problem, *the* text problem, which is the grand interplay of written materials, their interconnections, and the minds that play on these interconnections like harp-strings.

There is a specialty in the computer trade called "system design." Now, we all design systems, but is their a right way? (In some cases a system is intended to do something utterly new, and so there is nothing to be studied or replaced, but that is not germane to our problem.)

By some accounts, system design is the study of existing methods in some area of human endeavor, and the translation of these methods to a computer equivalent of some kind. But this broad description is not very helpful.

To observe and copy is not enough. True, the job of the system designer is in part to observe what people do in an existing system and take note of all the different activities that he sees-- no matter what the people think they are doing or seem to be doing. But the job of the actual design requires more. A designer necessarily makes compressions and adaptations. This is the creative part. And the designer should seek simplicity.

There is no reason, in systems design, to ratify and perpetuate those individual and local complications of life which have arisen in various contexts. Just as the scientist seeks generalizations, unifying ideas which summarize and compress all the varying details he may observe,; so the system designer seeks also an underlying structure in whatever existing system he is adapting to the computer. But this structure is not merely empirically found; in part it is imaginatively created, and represents the designer's conceptual encapsulation of what he thinks is going on, and what he thinks should go on.