9 . 3 SGML: The Standard andHandbook
SUMMARY
This case study is about the production of two intimately related documents. The first document is SGML, the ISO standard itself. The second document is The SGML Handbook, a book about the ISO standard.
BACKGROUND
The product of any standardsmaking activity is a document. Since SGML is a standard for the electronic publishing industry, it was only fitting (and elegant) that the ISO standard document be produced with the aid of SGML. The SGML Handbook (hereafter called the Handbook) is a book published six years after the ISO standard that contains the entire text of the SGML standard. The Handbook, presents the SGML standard in two complete ways: in its original order with annotations and in a logically structured overview. In addition it presents SGML with tutorial material explaining basic and advanced concepts.
THE STANDARD
SGML started out as an ANSI project called "Computer Languages for the Processing of Text" back in 1978. IBM's Charles Goldfarb made a presentation to the project committee, which eventually changed focus to the development of a language for the description, rather than processing, of text. The project was eventually taken over by the ISO committee TC97/SC18/WG8.
The technical development and evolution of SGML was the result of the work of many people. However, during the entire development cycle, Charles Goldfarb was the principal designer and editor. Coordination of the document was handled through conventional mailings of printed output, phone conversations, and working group meetings. The actual document processing system used for most of the process was IBM's Document Composition Facility (DCF), which used GML (Generic Markup Language), the predecessor to SGML. GML provided a head start in the development of SGML, but it was by no means the same thing. Toward the end of the development process, it was possible for a conversion program to translate GML into SGML.
Approximately one year before completion of the standard, Goldfarb went to the ISO Central Secretariat to clarify some language issues and formatting concerns. This meeting laid the groundwork for changing the way ISO does business. Goldfarb, Anders Berglund, of CERN, and Keith Brannon, of ISO, clearly demonstrated the true value of the SGML approach. They formatted the markedup text of the standard in five different ways in a period of about 30 minutes without touching the contents. Today, ISO is committed to producing all its standards using SGML.
The famous proof-of-the-pudding SGML story goes roughly as follows: Goldfarb shipped the unformatted standard to Anders Berglund at ISO via a computer network (Internet). Three days later, Goldfarb boarded a plane to Geneva. Upon arriving at the Geneva Hilton, he was greeted with, "Oh, Dr. Goldfarb, there is a package here for you." It was the complete, SGML standard, formatted to ISO specifications.
THE HANDBOOK
Although the production of the SGML standard was an SGML first, many of the Handbook's formatting requirements were more demanding and interesting.
The SGML Handbook was published in 1990 by Oxford University Press. The books editing and formatting were done mainly by Yuri Rubinsky, the late president of SoftQuad Inc., a major SGML vendor. The Handbook is unusual in several ways. First and foremost, it contains the entire text of ISO 8879, the SGML standard.
Second, it contains a novel cross-referencing system, which consists of roughly 3000 visual "buttons" that look like 
and are quite effective. In the button illustrated here, the [128] refers to a syntax production (parts of the definition of the standard), the 410 is a page number, and the 17 is a line number of ISO text. ISO text is text that originated in the actual standard and is typographically distinct from the new material in the Handbook. The cross-referencing mechanism is made more usable by the book's physical construction, which includes two ribbon bookmarks. The original "syntax production with built in cross-reference" idea came from Harvey Bingham of Interleaf Corporation. David Slocombe of SoftQuad, Inc., improved it with formatting-for-readability of the syntax productions.
The integration of several ISOowned texts with new material is unique. The SGML standard was originally tagged with SGML, and the same tagged text was imported into this book.
A new DTD for the book was developed that contained all the structure used by the standard and the new structure needed. The ISO text was not modified in any way, with the exception of the addition of the cross-reference buttons. Most of these buttons were generated automatically from the SGML markup; Yuri Rubinsky the editor, added the rest.
LEGAL ISSUES
Standards are explicit codifications of technology or a set of knowledge. They are written with precision and are generally difficult to read. The strict legalese in which they are written is necessary because standards can become legally binding. In addition, the standards-making process forces a certain structure and style upon the document. The Handbook is a reorganized version of SGML with readable explanations and commentary throughout. The primary contribution of this book is to make a difficult topic accessible.
The Handbook includes three sources of ISOowned information, which are interwoven with new material.
1. The entire text of ISO 8879.
2. The 1988 amendment to ISO 8879.
3. The document "ISO/IEC/JTC1/SC18/WG8/N1035: Recommendations for a Possible Revision of ISO 8879."
From a legal point of view, the text of an ISO standard is copyrighted by ISO. The legal nature and use of standards require that ISO be protect its standards to ensure that no unauthorized changes of content or intent are introduced. ISO also must continue to obtain revenue from the sale of standards. Like all organizations it must have resources.
Two major legal issues were resolved to bring the book into existence. First ISO receives a significant royalty from Handbook sales. This seems reasonable; since the book contains the entire text and commentary, why would anyone buy a copy of the standard directly from ISO? Second, and technically much more interesting, is the use and formatting of the three ISO documents. ISO had to be convinced that the content of the standard was not going to be modified in any way when it appeared in the Handbook. Technically, this was accomplished by using the same SGML-marked-up electronic files used to create the standard.
The 1988 amendment to the standard, which contains a small number of replacement sections, and the recommended changes from the N1035 document, are used in the structured overview and annotations. Change bars indicate that the material came from the amendment or N1035, rather than the 1986 standard.
FROM STANDARD TO HANDBOOK
The standard and much of the Handbook were keyed in using IBM's DCF system. A program written by Wayne Wohler of IBM was used to convert the original GML documents into SGML documents.
The Handbook was actually produced by Yuri Rubinsky. Rubinsky was sent dictation tapes for much of the new material for the Handbook. These were transcribed into SoftQuad's Author/Editor (an SGMLsensitive word processor). The resulting files, a collection of the original ISO 8879 standard, the amendment and N1035, and the new material were merged. The merged files were validated both by Goldfarb, using a parser he wrote during the development of the SGML, and by SoftQuad's SGML parser. This ensured that the files were conformant to SGML.
The final formatted version of the Handbook was produced using SoftQuad's Publishing Software. The publishing software provides the connection between the structural representation captured in SGML and the visual representation of the document. In fact this software converts SGML markup into troff (See Section 4 . 1 . 2 Language Characteristics in Chapter 4 Form and Function of Document Processors for more information on troff.)
CONCLUSIONS
The SGML Handbook sets a new standard for the publication of standards. One aspect of the Handbook is clear; it demonstrates the way all standards should be treated when printed. The republication of standards by knowledgeable participants should be encouraged by the standardsmaking organizations. ISO, Goldfarb, and Rubinsky all deserve congratulations for breaking new ground with publication of the The SGML Handbook.
