2. How are man pages accessed?

You need to know the precise mechanism for acccessing man pages in order to give your man page the right name and install it in the right place. Each man page should be categorized in a specific section, denoted by a single character. The most common sections under Linux, and their human readable names, are:

Section The human readable name
   1    User commands that may be started by everyone.
   2    System calls, that is, functions provided by the kernel.
   3    Subroutines, that is, library functions.
   4    Devices, that is, special files in the /dev directory.
   5    File format descriptions, e.g. /etc/passwd.
   6    Games, self-explanatory.
   7    Miscellaneous, e.g. macro packages, conventions.
   8    System administration tools that only root can execute.
   9    Another (Linux specific) place for kernel routine documentation.
   n    (Deprecated) New documentation, that may be moved to a more appropriate section.
   o    (Deprecated) Old documentation, that may be kept for a grace period.
   l    (Deprecated) Local documentation referring to this particular system.

The name of the man page's source file (the input to the formatting system) is the name of the command, function or file name, followed by a dot, followed by the section character. If you write the documentation on the format of the `passwd' file you have to name the source file `passwd.5'. Here we also have an example of a file name that is the same as a command name. There might be even a library subroutine named passwd. Sectioning is the usual way to resolve these ambiguities: The command description is found in the file `passwd.1' and the hypothetical library subroutine in `passwd.3'.

Sometimes additional characters are appended and the file name looks for example like `xterm.1x' or `wish.1tk'. The intent is to indicate that this is documentation for an X Window program or a Tk application, respectively. Some manual browsers can make use of this additional information. For example xman will use `xterm(x)' and `wish(tk)' in the list of available documentation.

Please don't use the n, o and l sections; according to the File System Standard these sections are deprecated. Stick to the numeric sections. Beware of name clashes with existing programs, functions or file names. It is certainly a bad idea to write yet another editor and call it ed, sed (for smart ed) or red (for Rocky's ed). By making sure your program's name is unique, you avoid having someone execute your program but read someone else's man page, or vice versa.

Now we know the name to give our file. The next decision is the directory in which it will finally be installed (say, when the user runs `make install' for your package.) On Linux, all man pages are below directories listed in the environment variable MANPATH. The doc-related tools use MANPATH in the same way the shell uses PATH to locate executables. In fact, MANPATH has the same format as PATH. Each contains a colon-separated list of directories (with the exception that MANPATH does not allow empty fields and relative pathnames -- it uses absolute names only.) If MANPATH is not set or not exported, a default will be used that contains at least the /usr/man directory. To speed up the search and to keep directories small, the directories specified by MANPATH (the so-called base directories) contain a bunch of subdirectories named `man<s>' where <s> stands for the one-character section designator introduced in the table above. Not all of the sections may be represented by a subdirectory because there simply is no reason to keep an empty `mano' subdirectory. However, there may be directories named `cat<s>', `dvi<s>' and `ps<s>' which hold documentation that is ready to display or print. More on this later. The only other file in any base directory should be a file named `whatis'. The purpose and creation of this file will also be described under paragraph 12). The safest way to have a man page for section <s> installed in the right place is to put it in the directory /usr/man/man<s>. A good Makefile, however, will allow the user to chose a base directory, by means of a make variable, MANDIR, say. Most of the GNU packages can be configured with the --prefix=/what/ever option. The manuals will then be installed under the base directory /what/ever/man. I suggest you also provide a way to do something similar.

With the advent of the Linux File System Standard (FS-Stnd), things became more complicated. [Note: the FS-Stnd appears to be replaced by the Filesystem Hierarchy Standard, FHS.] The FS-Stnd 1.2 states that

"Provisions must be made in the structure of /usr/man to support manual pages which are written in different (or multiple) languages."

This is achieved by introducing another directory level that distinguishes between different languages. Quoting again from FS-Stnd 1.2:

"This naming of language subdirectories of /usr/man is based on Appendix E of the POSIX 1003.1 standard which describes the locale identification string -- the most well accepted method to describe a cultural environment. The <locale> string is: <language>[_<territory>][.<character-set>][,<version>]"

(See the FS-Stnd for a few common <locale> strings.) According to these guidelines, we have our man pages in /usr/man/<locale>/man[1-9lno]. The formatted versions should then be in /usr/man/<locale>/cat[1-9lno] of course, otherwise we could only provide them for a single locale. HOWEVER, I can not recommend switching to that structure at this time. The FS-Stnd 1.2 also allows that

"Systems which use a unique language and code set for all manual pages may omit the <locale> substring and store all manual pages in <mandir>. For example, systems which only have English manual pages coded with ASCII, may store manual pages (the man[1-9] directories) directly in /usr/man. (That is the traditional circumstance and arrangement in fact.)"

I would not switch until all tools (like xman, tkman, info and many others that read man pages) can cope with the new structure.