As Linux becomes more and more popular, a lot of documentation is required, not only for newcomers, but for all users. Just think of all the FAQs, HOWTOs, manual pages, and books everyone needs for their daily work. Some people want to read these documents as plain ASCII text, while others want to read them over the World Wide Web or print them on their PostScript printer. It is possible to make an HTML version of an ASCII document for the Web and a nicely-formated PostScript version for people to print, but all the different formats have to be maintained separately. This is theoretically possible, but doesn't happen in real life.

We need a documentation system that can produce different formats from a single source. The Linux Documentation Project faced this exact dilemma when the HOWTO project was started, so Matt Welsh wrote the Linuxdoc-SGML package to solve it. With this package, all documentation is formatted in a similar way. But SGML is very flexible, so you can use the system to write many different kinds of documentation; as an example, the XFree86 project uses Linuxdoc-SGML for all of its documentation.

<!doctype linuxdoc system>
<article>
<title>The Very Short Story
<author>A. Author
<date>1 Jan 1970
<p>
Once upon a time, they lived happily ever after.
<article>

As you can see here, the Linuxdoc-SGML syntax is very simple. Commands are written in angle brackets: <command>. When they apply to a block of the text they appear in a pair surrounding that block, so </article> before the block is balanced by </article> after the block. There is also an abbreviation for the latter case if the block is short: <tt/typewriter font/.

The first line of the document specifies the document type. Here you will always specify linuxdoc system, since this refers to the main macros of the Linuxdoc-SGML package. Then you start your document with the <article> command and close it at the end with the corresponding “article off” command </article>. The article itself starts with the title, the author, and the date (which is optional). After that you can start writing the body text. The <p> command indicates the beginning of the first paragraph. You don't have to worry about spaces or line breaks when writing the text, since multiple spaces between words are ignored and line breaks are automatically inserted at the appropriate positions. To begin a new paragraph, insert a blank line, which is a “synonym” for <p>.

Linuxdoc-SGML is actually a collection of programs that work together to provide the final output. You need to know how to use each of them; several examples will help. The format program creates files designed for LaTeX, groff, or makeinfo, and is part of the process of creating HTML files, which is explained more fully below. The -T argument tells format which program it is writing files for.

There is one utility for running each of the formatting programs (groff, etc). Each has a name starting with “q”, like “qtex”.

To get a PostScript file via LaTeX, just type

format -T latex example.sgml > example.tex
qtex example

and Linuxdoc-SGML will create a LaTeX-format file, use LaTeX to process that file, then use dvips to turn that into the PostScript file example.ps. Note that you need to have LaTeX and dvips installed, along with Linuxdoc-SGML, for this to work.

If you prefer a DVI file, you may use a -d switch with qtex:

format -T latex example.sgml | qtex -d > example.dvi

The plain ASCII output is created with a similar procedure. Just run:

format -T nroff example.sgml | qroff > example.txt

To get texinfo output that can be read with the GNU info program, use:

format -T info example.sgml

This will create the necessary files in the current directory automatically. Of course, you need the GNU texinfo package installed on your system to make texinfo files.

The HTML output needs a little bit more care, since two compilation stages are necessary to get all cross references built. First, you have to have the LINUXDOC environment variable set up correctly; you will want to put a line such as:

export LINUXDOC=~/linuxdoc-sgml-1.2

in your bash startup file, or:

setenv LINUXDOC=~/linuxdoc-sgml-1.2

in your csh or tcsh startup file.

Once that is working, you have to run several commands to get finished HTML:

format -T html example.sgml | prehtml | \
   fixref > tmp.html
format -T html example.sgml | prehtml >> tmp.html
cat tmp.html | html2html example > example.html
rm tmp.html

It's a good idea to put these commands in a shell script since you will call these commands often. Here's a simple version you can use:

bin/bash
bin/bash
# sgml2html
[ -z -$1- ] && { echo -What file?-; exit 1 }
BASE=`basename $1 .sgml'
[ ! -f $BASE.sgml ] && { echo -No file $BASE.sgml-; exit 2 }
TMP=$$tmp.html
format -T html $1 | prehtml | fixref > $TMP
format -T html $1 | prehtml >> $TMP
cat $TMP | html2html $BASE > $BASE.html
rm $TMP

This script requires that your input file has the extension .sgml.

This script must be given the full file name, and it requires that the file have the extension .sgml to work correctly.

All of this is documented more completely in the excellent, short manual provided with Linuxdoc-SGML.

Now that we have explained the necessary “overhead”, we can start writing larger documents. Therefore, you will probably want to create sections and subsections, start the document with an abstract, and insert a table of contents. You can just add the following lines after the date line:

<abstract>
Here is the abstract.
<abstract>
<toc>
<sect>The First Section
<sect1>The First Subsection

Depending on the output format you want to create, the table of contents will have different styles: in a LaTeX document you get a standard table of contents, and in an HTML file you will get a list of cross references for the different sections and subsections.

As you can see from our example, the <sect> command creates a new section and the <sect1> command a new subsection. You can access five different levels simply by increasing the number in the command name. Note that you will still have to insert the <p> command to start the first paragraph after an sectioning command. For technical documentation it is often necessary to include “verbatim” text—text that is passed through without interpretation by the SGML parser. You can do this with the <verb> command:

<verb>This text is not interpreted <verb>

However, some very special characters do need some additional handling even when they appear inside a verbatim environment. For example, the sequence of an opening angle bracket and a slash (</) has to be written as &ero;etago;. The manual that comes with the Linuxdoc-SGML package includes a list of all special characters and the commands used to enter them as literal characters.

The Linuxdoc-SGML package supports three different font styles beside the normal font: boldface, italics, and typewriter. To select a font insert the command bf, em, or tt, respectively. To switch back to the default font just use the corresponding “slashed” command, as in the following example:

This <em>text</em> is written in italics!

An abbreviated version of each of these is sometimes useful:

This <em/text/ is written in italics!

Now let's take a look at different list types that are supported:

You just start the list “environment” with the <list> command and close it with </list>. To produce a new item (either a bullet or a number) you can use the <item> command. In the description environment, you have to use the <tag> command with an argument that contains the “keyword”, as in the following example:

<descrip>
<tag/First./ This is the first point.
<tag>Second.</tag> And this is the second.
</descrip>

If you want to create WWW pages with this package, you may want to create cross references to other WWW pages. This is done with the <url> command, as in the following example:

<url url=-https://gnus.com/pub/text.html- name=-Text
document->

This creates a link to the WWW page text.html on the specified host and displays Text document as the title of the link. When you create non-HTML documents, both the name and the URL are shown, so that the reader can still find the document.