etc: Added an explicit .wsgi2 extension in the Python configuration.

[ashd.git] / doc / dirplex.doc

dirplex(1)
==========

NAME
----
dirplex - Physical directory handler for ashd(7)

SYNOPSIS
--------
*dirplex* [*-hN*] [*-c* 'CONFIG'] 'DIR'

DESCRIPTION
-----------

The *dirplex* handler maps URLs into physical files or directories,
and, having found a matching file or directory, it performs various
kinds of pattern-matching against its physical name to determine what
handler to call in order to serve the request. The mapping procedure
and pattern matching are described below.

Having found a handler to serve a file or directory with, *dirplex*
adds the `X-Ash-File` header to the request with a path to the
physical file, before passing the request on to the handler.

*dirplex* is a persistent handler, as defined in *ashd*(7).

OPTIONS
-------

*-h*::

	Print a brief help message to standard output and exit.

*-N*::

	Do not read the global configuration file `dirplex.rc`.

*-c* 'CONFIG'::

	Read an extra configuration file. If 'CONFIG' contains any
	slashes, it is opened by that exact name. Otherwise, it is
	searched for in the same way as the global configuration file
	(see CONFIGURATION below).

URL-TO-FILE MAPPING
-------------------

Mapping URLs into physical files is an iterative procedure, each step
looking in one single physical directory, starting with 'DIR'. For
each step, a path element is stripped off the beginning of the rest
string and examined, the path element being either the leading part of
the rest string up until (but not including) the first slash, or the
entire rest string if it contains no slashes. If the rest string is
empty, the directory being examined is considered the result of the
mapping. Otherwise, any escape sequences in the path element under
consideration are unescaped before examining it.

If the path element names a directory in the current directory, the
procedure continues in that directory, unless there is nothing left of
the rest string, in which case *dirplex* responds with a HTTP 301
redirect to the same URL, but ending with a slash. Otherwise, the
remaining rest string begins with a slash, which is stripped off
before continuing. If the path element names a file, that file is
considered the result of the mapping (even if the rest string has not
been exhausted yet).

If the path element does not name anything in the directory under
consideration, but contains no dots, then the directory is searched
for a file whose name before the first dot matches the path
element. If there is such a file, it is considered the result of the
mapping.

If the result of the mapping procedure is a directory, it is checked
for the presence of a filed named by the *index-file* configuration
directive (see CONFIGURATION below). If there is such a file, it is
considered the final result instead of the directory itself. If the
index file name contains no dots and there is no exact match, then,
again, the directory is searched for a file whose name before the
first dot matches the index file name.

See also 404 RESPONSES below.

CONFIGURATION
-------------

Configuration in *dirplex* comes from several sources. When *dirplex*
starts, unless the *-N* option is given, it tries to find a global
configuration file named `dirplex.rc`. It looks in `$HOME/.ashd/etc`,
and then in all directories named by the *PATH* environment variable,
appended with `../etc/ashd`. For example, then, if *PATH* is
`/usr/local/bin:/bin:/usr/bin`, the directories `$HOME/.ashd/etc`,
`/usr/local/etc/ashd`, `/etc/ashd` and `/usr/etc/ashd` are searched
for `dirplex.rc`, in that order. Only the first file found is used,
should there exist several.

If the *-c* option is given to *dirplex*, it too specifies a
configuration file to load. If the name given contains any slashes, it
is opened by that exact name. Otherwise, it is searched for in the
same manner as the global configuration file.

In addition, all directories traversed by *dirplex* when mapping a URL
into a physical file may contain a file called `.htrc`, which may
specify extra configuration options for all files in and beneath that
directory.

`.htrc` files are checked periodically and reread if changed. The
global configuration file and any file named by the *-c* option,
however, are never reexamined.

When using the configuration files for deciding what to do with a
found file, they are examined in order of their "distance" from that
file. `.htrc` files found in the directory or directories containing
the file are considered "closest" to the file under consideration,
followed by any configuration file named by the *-c* option, followed
by the global configuration file.

Each configuration file is a sequence of configuration stanzas, each
stanza being an unindented starting line, followed by zero or more
indented follow-up lines adding options to the stanza. The starting
line of a stanza is referred to as a "configuration directive"
below. Each line is a sequence of whitespace-separated words. A word
may contain whitespace if such whitespace is escaped, either by
enclosing the word in double quotes, or by escaping individual
whitespace characters with a preceding backslash. Backslash quoting
may also be used to treat double quotes or another backslash literally
as part of the word. Empty lines are ignored, and lines whose first
character after leading whitespace is a hash character (`#`) are
treated as comments and ignored.

The following configuration directives are recognized:

*include* ['FILENAME'...]::

	Read the named files and act as if their contents stood in
	place of the *include* stanza. A 'FILENAME' may be a glob
	pattern, in which case all matching files are used, sorted by
	their filenames. If a 'FILENAME' is a relative path, it is
	treated relative to the directory containing the file from
	which the *include* stanza was read, even if the inclusion has
	been nested. Inclusions may be nested to any level.

*index-file* ['FILENAME'...]::

	The given 'FILENAMEs' are used for finding index files (see
	URL-TO-FILE MAPPING above). Specifying *index-file* overrides
	entirely any previous specification in a more distant
	configuration file, rather than adding to it. Zero 'FILENAMEs'
	may be given to turn off index file searching completely. The
	*index-file* directive accepts no follow-up lines.

*child* 'NAME'::

	Declares a named, persistent request handler (see *ashd*(7)
	for a more detailed description of persistent handlers). It
	must contain exactly one follow-up line, *exec* 'PROGRAM'
	['ARGS'...], specifying the program to execute and the
	arguments to pass it. If given in a `.htrc` file, the program
	will be started in the same directory as the `.htrc` file
	itself. The *child* stanza itself serves as the identity of
	the forked process -- only one child process will be forked
	per stanza, and if that child process exits, it will be
	restarted the next time the stanza would be used. If a `.htrc`
	file containing *child* stanzas is reloaded, any currently
	running children are reused for *child* stanzas in the new
	file with matching names (even if the *exec* line has
	changed).

*fchild* 'NAME'::

	Declares a named, transient request handler (see *ashd*(7) for
	a more detailed description of transient handlers). It must
	contain exactly one follow-up line, *exec* 'PROGRAM'
	['ARGS'...], specifying the program to execute and the
	arguments to pass it. In addition to the specified arguments,
	the HTTP method, raw URL and the rest string will be appended
	as described in *ashd*(7). If given in a `.htrc` file, the
	program will be started in the same directory as the `.htrc`
	file itself.

*match* [*directory*]::

	Specifies a filename pattern-matching rule. The
	pattern-matching procedure and the follow-up lines accepted by
	this stanza are described below, under MATCHING.

*capture* 'HANDLER' ['FLAGS']::

	Only meaningful in `.htrc` files. If a *capture* directive is
	specified, then the URL-to-file mapping procedure as described
	above is aborted as soon as the directory containing the
	`.htrc` file is encountered. The request is passed, with any
	remaining rest string, to the specified 'HANDLER', which must
	be a named request handler specified either in the same
	`.htrc` file or elsewhere. The *capture* directive accepts no
	follow-up lines. Note that the `X-Ash-File` header is not
	added to requests passed via *capture* directives.  If 'FLAGS'
	contain the character `R`, this *capture* directive will be
	ignored if it is in the root directory that *dirplex* serves.

MATCHING
--------

When a file or directory has been found by the mapping procedure (see
URL-TO-FILE MAPPING above), the name of the physical file is examined
to determine a request handler to pass the request to. Note that only
the physical file name is ever considered; any logical request
parameters such as the request URL or the rest string are entirely
ignored.

To match a file, any *match* stanzas specified by any `.htrc` file or
in the global configuration files are searched in order of their
"distance" (see CONFIGURATION above) from the actual file. If it is a
directory which is being considered, only *match* stanzas with the
*directory* parameter are considered; otherwise, if it is a file, only
*match* stanzas without the *directory* parameter are considered.

A *match* stanza must contain at least one follow-up line specifying
match rules. All rules must match for the stanza as a whole to match.
The following rules are recognized:

*filename* 'PATTERN'...::

	Matches if the name of the file under consideration matches
	any of the 'PATTERNs'. A 'PATTERN' is an ordinary glob
	pattern, such as `*.php`. See *fnmatch*(3) for more
	information.

*pathname* 'PATTERN'...::

	Matches if the entire path of the file under consideration
	matches any of the 'PATTERNs'. A 'PATTERN' is an ordinary glob
	pattern, except that slashes are not matched by wildcards. See
	*fnmatch*(3) for more information. If a *pathname* rule is
	specified in a `.htrc` file, the path will be examined as
	relative to the directory containing the `.htrc` file, rather
	than to the root directory being served.

*default*::

	Matches if and only if no *match* stanza without a *default*
	rule matches (in any configuration file).

*local*::

	Valid only in `.htrc` files, *local* matches if and only if
	the file under consideration resides in the same directory as
	the containing `.htrc` file.

In addition to the rules, a *match* stanza must contain exactly one
follow-up line specifying the action to take if it matches. The
following actions are recognized:

*handler* 'HANDLER'::

	'HANDLER' must be a named handler (see CONFIGURATION
	above). The named handler is searched for not only in the same
	configuration file as the *match* stanza, but in all
	configuration files that are valid for the file under
	consideration, in order of distance. As such, a more deeply
	nested `.htrc` file may override the specified handler without
	having to specify any new *match* stanzas.

*fork* 'PROGRAM' ['ARGS'...]::

	Run a transient handler for this file, as if it were specified
	by a *fchild* stanza. This action exists mostly for
	convenience.

A *match* stanza may also contain any number of the following,
optional directives:

*set* 'HEADER' 'VALUE'::

	If the *match* stanza is selected as the match for a file, the
	named HTTP 'HEADER' in the request is set to 'VALUE' before
	passing the request on to the specified handler.

*xset* 'HEADER' 'VALUE'::

	*xset* does exactly the same thing as *set*, except that
         'HEADER' is automatically prepended with the `X-Ash-`
         prefix. The intention is only to make configuration files
         look nicer in this very common case.

404 RESPONSES
-------------

A HTTP 404 response is sent to the client if

 * The mapping procedure fails to find a matching physical file;
 * A path element is encountered during mapping which, after URL
   unescaping, either begins with a dot or contains slashes;
 * The mapping procedure finds a file which is neither a directory nor
   a regular file (or a symbolic link to any of the same);
 * An empty, non-final path element is encountered during mapping; or
 * The mapping procedure results in a file which is not matched by any
   *match* stanza.

By default, *dirplex* will send a built-in 404 response, but any
`.htrc` file or global configuration may define a request handler
named `.notfound` to customize the behavior. Note that, unlike
successful requests, such a handler will not be passed the
`X-Ash-File` header.

The built-in `.notfound` handler can also be used in *match* or
*capture* stanzas (for example, to restrict access to certain files or
directories).

EXAMPLES
--------

The *sendfile*(1) program can be used to serve HTML files as follows.

--------
fchild send
  exec sendfile

match
  filename *.html *.htm
  xset content-type text/html
  handler send
--------

Assuming the PHP CGI interpreter is installed on the system, PHP
scripts can be used with the following configuration, using the
*callcgi*(1) program.

--------
# To use plain CGI, which uses more resources per handled request,
# but less static resources:
fchild php
  exec callcgi -p php-cgi

# To use FastCGI, which keeps PHP running at all times, but uses less
# resources per handled request:
child php
  exec callfcgi multifscgi 5 php-cgi

match
  filename *.php
  handler php
--------

If there is a directory without an index file, a file listing can be
automatically generated by the *htls*(1) program as follows.

--------
match directory
  default
  fork htls
--------

The following configuration can be placed in a `.htrc` file in order
to dedicate the directory containing that file to some external SCGI
script engine. Note that *callscgi*, and therefore the script engine
itself, is started in the same directory, so that arbitrary code
modules or data files can be put directly in that directory and be
easily found.

--------
child foo
  exec callscgi scgi-wsgi -p . foo

capture foo
--------

AUTHOR
------
Fredrik Tolf <fredrik@dolda2000.com>

SEE ALSO
--------
*ashd*(7)