THE XANADU HYPERTEXT SYSTEM:

A NEW DOCUMENTATION MANAGEMENT TOOL

INTRODUCTION

A new documentation management tool, "hypertext", lets the user, rather than the software system, decide how information will be interrelated and stored. It is a system designed to handle text in all of it's complexity. Today's technical communicator faces information management nightmares such as changing one basic manual to reflect the differences in several installations, or to suit each particular customer's needs. Hypertext offers a unified way to solve such problems.

Currently, most documentation management consists of two phases. First, a document is written or edited using a text editor or word processing software package. It is then kept track of using storage and retrieval systems such as database management systems, keyword systems, or fulltext search systems. These systems implicitly determine the interconnections and comparisons a user can make between documents, whereas hypertext systems allow the user to make explicit interconnections called "links". To understand why hypertext systems are a unique concept, let us look more closely at word processing and some of the more popular storage and retrieval methods. A LOOK AT CURRENT DOCUMENTATION MANAGEMENT SYSTEMS Word processors and text editors are fine tools for manipulating text within documents, but are usually not well coordinated with storage and retrieval functions. They generally do not deal with relationships between documents that may exist. As Ted Nelson says (1), this state of affairs forces the writer or editor to stop working at the boundaries of a document, which, to some extent, limits thought. There is no convenient way to show relations between documents. Also, word processors and text editors often tend to limit the size of the documents a user can work with. Storing documents in files means that the user has to think up file names, worry about file conventions, and keep track of files on paper, or with a separate system. Most text editors or word processors have no way to keep track of changes resulting from the editing process other than keeping an old copy and a new copy online or a trail of all the editing changes. This means storing a lot of copies. In keyword systems, keywords are usually attached to a text for reference. A user asks the system to find all the articles with certain keywords contained within them, and it searches thru its database looking for them. Keyword systems function well as card catalogs; however, since not all thoughts can be expressed in one word, problems arise. What seems an obvious keyword to you may not be a keyword to the author. Thus the article is not listed by the word that you may be looking for, though it may be listed by keywords that you have not thought of. Keyword systems are not designed to let a user edit while remaining within the system. Explicit citations within or between documents cannot be created.

Keyword systems are a subset of the more general type of storage and retreival system called fulltext search systems. With fulltext search systems, a user types something like a phrase or a sentence which might be found in the document he or she wishes to find. The system searches its database for the text until the desired document is located. Fulltext search systems give users more latitude in expressing thought than keyword systems, thereby increasing the user's chances of finding all the pertinent references. But like keyword systems, fulltext search systems do not let users edit without exiting from the system. They do not allow users to make explicit citations, notes, or other connections between documents.

Data base management systems, or DBMS, provide excellent means of storing and retrieving any type of data which will fit into a fixed field format. Any data that can be grouped into lists may be stored in a database management system, which will let users make certain kinds of very useful comparisons between the fixed fields. Parts lists, numerical data, and personnel records are examples of data best suited to these systems. Data base management systems are not optimized to edit, store, or retrieve documents because text is too complicated. DBMS are sometimes used for keeping track of names of journals or articles. Because there is often no standard way of abbreviating journal names, users may have problems with this application.

The following table, Table 1, summarizes the differences between current documentation management tools and the Xanadu Hypertext System. HYPERTEXT: NONSEQUENTIAL WRITING

As the above discussion shows, the current approach to documentation management requires more than one tool to work with text and is restrictive when users need to interrelate documents. However, the real problem is that current systems are not designed to let people use computers in the same way and with the same ease that they use paper. Hypertext, which means nonsequential writing, is a system based on the way we use paper to read, write and file. The term was coined by Nelson, the originator of the concept (2,3).

To understand what nonsequential writing means, let us look at an example that will be familiar to many technical communicators. Consider the amount and the complexity of paperwork required to build an airplane.

Suppose that the project engineer decides to move the wings on an airplane to a slightly different position on the fuselage. Many design documents will be affected by this change, perhaps thousands of them. To change the wing, he must first find all the components that will need to be modified. The change must then be made, but only after searching through the blueprints to make sure that this change will not adversely affect anything else. Once this is done, the technical communicators must update all affected documents to reflect the changes.

For something as seemingly trivial as changing the position of one bolt, the engineer has to know what components may be affected. For example, rewiring may be required if the bolt is moved into the path of a cable. Since our fictitious aerospace company doesn't have strings connecting its technical documents together, the writer or someone in the project must spend an appreciable amount of time chasing down all the documents which refer to that bolt. A common expression in the aerospace industry is, "When the paperwork weighs as much as the aircraft, the plane is finished."

In the above example all documents (including blueprints) written about that plane constitute a literature, or a hypertext. Because everything contained within this literature concerns some aspect of the airplane, it is interconnected, as if by a network of invisible strings. DEVELOPMENT OF HYPERTEXT

The idea of hypertext was born in 1960 when Ted Nelson decided to design a computer system which would allow him to make multiple versions of articles or documents out of the same pieces of text. His ideas on hypertext and on what computers could be like were published in the 1974 underground classic

Computer Lib/Dream Machines.

The book attracted the attention of some brilliant young men who joined Nelson. They finished designing and began implementing a hypertext system that could retrieve hypertext in a reasonable response time out of a database (or "docuverse") as large as a world library (4). The effort, named Project Xanadu, created the Xanadu(tm) Hypertext System. It licensed a company, XOC, to market the system for commercial applications.

Douglas Engelbart and Andries van Dam have also done major work in the field of hypertext. Douglas Engelbart began work on an interactve editor and form of hypertext called NLS in 1959 (5). He is considered to be the father of several concepts related to hypertext such as word processing, networking, and teleconferencing. Now at TymeShare, his system is currently in use over a worlwide network.. The system is now called "Augment". It differs from Xanadu Hypertext in several aspects, the most notable being that links are embedded in the text and it is only capable of functioning with limited databases.

Andries van Dam implemented the first hypertext system in conjunction with Nelson at Brown University in 1968 (6). Since then two other hypertext systems have been implemented at Brown (7). Other hypertext research involved a prolonged use of hypertext in a poetry class (8).

I will be discussing hypertext as it is being implemented by XOC for the rest of this paper both because I am most familiar with it and because it is the most ambitious hypertext project to date. For a more complete discussion of other forms of hypertext mentioned see the excellent review article on interactive editing systems by Meyrowitz and van Dam (9).

THE XANADU HYPERTEXT SYSTEM

My colleagues at XOC who designed the Xanadu Hypertext System wanted it to be so simple that a child (or an English major!) could use it. Therefore they createda system design that lets us hide all the arcane and proprietary inner workings from users. Our system has two parts- a "backend" that the user never sees, and a "frontend" that the user sees and uses. Don't be misled, however; this does not mean that the user sees two parts. It simply refers to internal organization of functions within a unified system.

The backend is responsible for storage and retrieval functions. This part of the system runs on large or very powerful computers which can be networked together. Every Xanadu Hypertext System will have the same type of backend.

The frontend is responsible for display function and user interaction, including editing. Frontend software will be modified to fit onto various types and sizes of computers, including personal computers. Since users and their needs differ, our designers wanted to make sure that the part of the system that they use may be modified to fit their tastes and preferences.

The Xanadu system is built out of "documents" and "links". A document is defined to be a string of contiguous characters. Thus a document may be text, graphics, blueprints, numbers, data, digitized voice, or music. Text that a user may want to store and link on a Xanadu could be in the form of memos, technical documentation, business letters, marginal notes, citations, grafitti, or hypertext poetry. The contents of documents and the connections between them are limited only be a user's creativity. These explicit connections within a document or between documents are called "links". Arbitrary pieces of arbitrary documents and links themselves may be linked together. Links can be retrieved at will, along with information about what is being linked to what and about the nature and type of the link itself. A link is bidirectional so you can tell which point it came from and where it is going. The system can correctly maintain these links in the midst of changes to the very data being interlinked. There are different link types for different purposes. Although just how a link type is displayed depends on which frontend is used, the general forms can be discussed. Jumplinks are used to follow citations. In the current experimental frontend, a user sees jumplinks as reverse video areas in the text he or she is reading. To find the other end of a jumplink or any other link type, the user merely hits a control key on the computer keyboard. A window large enough to display the cited text opens within the original document so that the user may read the pertinent material, then continue reading the original text. If the new material proves to be more interesting, the user can enlarge the window and read the new document rather than the original.

Say that a user begins reading a document, and that he or she sees an interesting looking jump link to follow. After following the jump link, the user sees a second interesting citation within the windowed document. The user may "pass through the window" and call up a third document. If the third document proves to be uninteresting, the user may follow the same links back to his or her original document. Xanadu links allow users to pass through an infinite number of windows and back again (only a certain number of windows will be displayed on the screen at once). Nelson refers to this action as "navigating through the docuverse" (10).

Other link types include quote links, footnote links, and marginnote links. They differ from jumplinks and each other only in the way in which they are indicated within a document, and where their type and size of display window will appear on the screen. Perhaps the most useful feature of Xanadu for writers is its versioning capacity. Xanadu hypertext can maintain multiple versions of any given document, efficiently storing the common portions in common, with separate portions of storage required only for those parts of the documents which are actually different.

The frontend will display a vertically split screen with a different version of the document shown on either side. Correspondences between versions will be indicated with connecting lines, and users will be able to scroll around in either version and edit.

All such editing wil be logged by Xanadu's historical trace function. Xanadu can provide historical traceback information in dated chronological order about any and all changes to a given document. It can do all of this efficiently, with only logarithmic overhead in terms of both storage space and response time. For example, if the number of stored documents increases geometrically, the response time required to locate a particular one will increase only linearly.

Historical trace is not yet implemented, but it might look something like a branching tree on a future frontend. A user could scroll along the tree with a mouse or a cursor until he or she reached the branch indicating the desired version of the document. A touch of a control key would display the document and associated traceback information on the screen. USES OF HYPERTEXT As the above discussion shows, the Xanadu hypertext system can help the technical communicator perform a variety of tasks with greater efficiency. The linkage and versioning capacities will make documentation updates less of a chore. Theversioning facility can greatly help those who need to edit previously written materials into new documentation. Links make it possible for communicators to easily keep track of their own and their departments' work, and to obtain feedback from users of online documentation. User feedback logs can be constructed by allowing users to link their own comments to text they find obscure or confusing. A unified information management system would also be useful to personnel outside the technical communications department. A hypertext system is such a general tool that it can be used for office automation, conferencing, and engineering projects. Hypertext is ideal for managers and executives who need a way to have iformation at their fingertips. It would enable them to make better decisions without wasting valuable time in meetings and papershuffling. As of this writing the Xanadu organization has a working hypertext demonstration and plans to deliver a working prototype system within the year. Different kinds of links work but the backend still requires six months to a year development work before the Xanadu Hypertext System is an "off the shelf product". Currently, the system runs on a Sun(trademark of Sun Microsystems) 68000 Workstation. It is written in the 'C' programming language running under the UNIX(trademark of Bell Laboratories) operating system, and can be made to work on almost any computer. When completed it will run on larger,modern machines with frontends running on a variety of machines, depending on demand.

SUMMARY In summary, a new, more powerful and unified documentation management system called the Xanadu Hypertext System will soon be available. It will give users the freedom to organize documents to fit their needs, rather than the computer's.

References

(1) Nelson, T.H., "Replacing the Printed Word", INFORMATION PROCESSING 80, S.H. Lavington, (ed.), North Holland Publishing Company, IFIP, 1980 pp. 1013-23

(2) Nelson, T.H.,

Computer Lib/Dream Machines,

The Distributors, South Bend, IN., 1974

(3),(4),(10) Nelson, T.H.,

Literary Machines,

T.H. Nelson, Swarthmore, PA., 1981

(5) Engelbart, D.C., "The augmented knowledge workship" in Proc. National Computer Conf., vol. 42, AFIPS Press, Arlington, VA., 1973, pp.9-21

(6)Carmody, S., Gross, W., Nelson, T.H., Rice, D.E., and van Dam, A. "A hypertext editing concept for the /360," in Pertinent Concepts in Computer Graphics, M. Faiman and Nievergelt, J., Eds., University of Illinois Press, Urbana, Ill., 1969 pp. 291-330

(7),(9) Meyrowitz, N., and van Dam, A. "Interactive Editing Systems, parts I and II", Computing Surveys of the ACM 14 (1982), 321-353

(8) Catano, J. "Poetry and Computers: Experimenting with the Communal Text," Comput. Hum. 13 (1979), 269-275