Standards for Electronic Publishing: SGML and ANSI Z39.59
By Evan Owens
Journals Division, The University of Chicago Press

Until fairly recently, paper was the universal format for text
interchange between authors, editors, publishers, and readers.  As
we move into an electronic future, however, we find ourselves using
a plethora of file formats: typesetting systems (e.g., Penta,
XyVision, or TeX), proprietary word processors (e.g., WordPerfect
or Microsoft Word), text interchange formats (e.g., ASCII,
Rich-Text-Format, or Navy DIF), and markup languages (e.g., LaTeX
or SGML).  The choice of a standard format is critical because we
expect to use electronic files not just as a new way to produce the
paper journal but as the basis for new ways to store and distribute
texts electronically.  We have five years of the Astrophysical
Journal stored on magnetic tape in a proprietary typesetting
format; once a format is chosen, these tapes can be translated and
used as the basis of an electronic archive to which new material
can be added as it is published. 
     A consensus has emerged in the publishing community that the
foundation for electronic publishing will be the adoption of
Standard Generalized Markup Language (SGML) otherwise known as ISO
8879 (1986).  SGML was designed as a language for electronic
documents and is well suited for text databases, hypertext, CD-ROM,
and electronic books and journals.  SGML documents are not
dependent on any hardware, software, formatter, or operating
system.   The concept is important.  Generalized  markup is codes
or tags that describe the content of the text such as "heading" or
"title" or "footnote"; it is completely different from procedural
markup, codes that describe the format such as "10 point Times
Roman" or "indent 1 em."  SGML provides a way to describe and
validate the structure or hierarchy of a document (e.g., second
level headings go inside first level headings, front matter comes
before the body of the article, and so on) through a document type
definition (DTD).  This is important for text databases, because it
makes it possible to search selected portions of the text; for
example, the footnotes, display equations, headings, references,
etc. 
     To use SGML, one has to have a suitable SGML document type
definition (DTD).   There is an ANSI standard DTD for articles  and
books which resulted from a joint project of the Association of
American Publishers (AAP), the Council on Library Resources, and
other organizations; it also includes markup specifications for
mathematics and tables.  We are constructing a superset of the AAP
DTDs to use for the Astrophysical Journal and other AAS 
publications. 
     With a DTD available, one then creates documents using 
the  specified names to tag the various parts of the manuscript. 
Ideally, the author would create an SGML document directly;
however, SGML coding is tedious without special software.  There
are SGML-aware validating text editors available, some with WYSIWYG
equation and table editors, but they are currently very expensive
or not suited for general use.  The situation should change for the
better early next year, however, as new SGML products are expected
from WordPerfect, Frame Technology, and others. 
     Not everyone will want to write using an SGML editor, 
even when such tools are widely available.  Fortunately, it is
possible to translate from one format to another.  At the
University of Chicago Press we now accept files for some of our
journals in common word processor formats and then translate, edit,
code, and translate again to typesetting systems.  For the
Astrophysical Journal, we plan to accept manuscripts in the AAS 
version of LaTeX and translate to SGML; this translation works
particularly well because of the structured markup present in a
properly coded LaTeX manuscript.   Plain TeX files can not be
translated as they contain procedural rather than generalized
markup codes.  We will then supply SGML files to our typesetters
and get SGML files back for archival use.