Use @copying, @title, @subtitle, @author.

[bpt/emacs.git] / doc / misc / url.texi

\input texinfo
@setfilename ../../info/url
@settitle URL Programmer's Manual

@iftex
@c @finalout
@end iftex
@c @setchapternewpage odd
@c @smallbook

@tex
\overfullrule=0pt
%\global\baselineskip 30pt      % for printing in double space
@end tex
@dircategory World Wide Web
@dircategory GNU Emacs Lisp
@direntry
* URL: (url).                 URL loading package.
@end direntry

@copying
This file documents the URL loading package.

Copyright @copyright{} 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2002,
2004, 2005, 2006, 2007, 2008 Free Software Foundation, Inc.

@quotation
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License, Version 1.2 or
any later version published by the Free Software Foundation; with no
Invariant Sections, with no Front-Cover Texts, and with no Back-Cover
Texts.  A copy of the license is included in the section entitled ``GNU
Free Documentation License''.
@end quotation
@end copying

@c
@titlepage
@title URL Programmer's Manual
@subtitle First Edition, URL Version 2.0
@author William M. Perry @email{wmperry@@gnu.org}
@author David Love @email{fx@@gnu.org}
@page
@vskip 0pt plus 1filll
@insertcopying
@end titlepage

@page
@node Top
@top URL


@menu
* Getting Started::             Preparing your program to use URLs.
* Retrieving URLs::             How to use this package to retrieve a URL.
* Supported URL Types::         Descriptions of URL types currently supported.
* Defining New URLs::           How to define a URL loader for a new protocol.
* General Facilities::          URLs can be cached, accessed via a gateway
                                and tracked in a history list.
* Customization::               Variables you can alter.
* GNU Free Documentation License:: The license for this documentation.
* Function Index::
* Variable Index::
* Concept Index::
@end menu

@node Getting Started
@chapter Getting Started
@cindex URLs, definition
@cindex URIs

@dfn{Uniform Resource Locators} (URLs) are a specific form of
@dfn{Uniform Resource Identifiers} (URI) described in RFC 2396 which
updates RFC 1738 and RFC 1808.  RFC 2016 defines uniform resource
agents.

URIs have the form @var{scheme}:@var{scheme-specific-part}, where the
@var{scheme}s supported by this library are described below.
@xref{Supported URL Types}.

FTP, NFS, HTTP, HTTPS, @code{rlogin}, @code{telnet}, tn3270,
IRC and gopher URLs all have the form

@example
@var{scheme}://@r{[}@var{userinfo}@@@r{]}@var{hostname}@r{[}:@var{port}@r{]}@r{[}/@var{path}@r{]}
@end example
@noindent
where @samp{@r{[}} and @samp{@r{]}} delimit optional parts.
@var{userinfo} sometimes takes the form @var{username}:@var{password}
but you should beware of the security risks of sending cleartext
passwords.  @var{hostname} may be a domain name or a dotted decimal
address.  If the @samp{:@var{port}} is omitted then the library will
use the `well known' port for that service when accessing URLs.  With
the possible exception of @code{telnet}, it is rare for ports to be
specified, and it is possible using a non-standard port may have
undesired consequences if a different service is listening on that
port (e.g., an HTTP URL specifying the SMTP port can cause mail to be
sent). @c , but @xref{Other Variables, url-bad-port-list}.
The meaning of the @var{path} component depends on the service.

@menu
* Configuration::
* Parsed URLs::                 URLs are parsed into vector structures.
@end menu

@node Configuration
@section Configuration

@defvar url-configuration-directory
@cindex @file{~/.url}
@cindex configuration files
The directory in which URL configuration files, the cache etc.,
reside.  Default @file{~/.url}.
@end defvar

@node Parsed URLs
@section Parsed URLs
@cindex parsed URLs
The library functions typically operate on @dfn{parsed} versions of
URLs.  These are actually vectors of the form:

@example
[@var{type} @var{user} @var{password} @var{host} @var{port} @var{file} @var{target} @var{attributes} @var{full}]
@end example

@noindent where
@table @var
@item type
is the type of the URL scheme, e.g., @code{http}
@item user
is the username associated with it, or @code{nil};
@item password
is the user password associated with it, or @code{nil};
@item host
is the host name associated with it, or @code{nil};
@item port
is the port number associated with it, or @code{nil};
@item file
is the `file' part of it, or @code{nil}.  This doesn't necessarily
actually refer to a file;
@item target
is the target part, or @code{nil};
@item attributes
is the attributes associated with it, or @code{nil};
@item full
is @code{t} for a fully-specified URL, with a host part indicated by
@samp{//} after the scheme part.
@end table

@findex url-type
@findex url-user
@findex url-password
@findex url-host
@findex url-port
@findex url-file
@findex url-target
@findex url-attributes
@findex url-full
@findex url-set-type
@findex url-set-user
@findex url-set-password
@findex url-set-host
@findex url-set-port
@findex url-set-file
@findex url-set-target
@findex url-set-attributes
@findex url-set-full
These attributes have accessors named @code{url-@var{part}}, where
@var{part} is the name of one of the elements above, e.g.,
@code{url-host}.  Similarly, there are setters of the form
@code{url-set-@var{part}}.

There are functions for parsing and unparsing between the string and
vector forms.

@defun url-generic-parse-url url
Return a parsed version of the string @var{url}.
@end defun

@defun url-recreate-url url
@cindex unparsing URLs
Recreates a URL string from the parsed @var{url}.
@end defun

@node Retrieving URLs
@chapter Retrieving URLs

@defun url-retrieve-synchronously url
Retrieve @var{url} synchronously and return a buffer containing the
data.  @var{url} is either a string or a parsed URL structure.  Return
@code{nil} if there are no data associated with it (the case for dired,
info, or mailto URLs that need no further processing).
@end defun

@defun url-retrieve url callback &optional cbargs
Retrieve @var{url} asynchronously and call @var{callback} with args
@var{cbargs} when finished.  The callback is called when the object
has been completely retrieved, with the current buffer containing the
object and any MIME headers associated with it.  @var{url} is either a
string or a parsed URL structure.  Returns the buffer @var{url} will
load into, or @code{nil} if the process has already completed.
@end defun

@node Supported URL Types
@chapter Supported URL Types

@menu
* http/https::                  Hypertext Transfer Protocol.
* file/ftp::                    Local files and FTP archives.
* info::                        Emacs `Info' pages.
* mailto::                      Sending email.
* news/nntp/snews::             Usenet news.
* rlogin/telnet/tn3270::        Remote host connectivity.
* irc::                         Internet Relay Chat.
* data::                        Embedded data URLs.
* nfs::                         Networked File System
@c * finger::
@c * gopher::
@c * netrek::
@c * prospero::
* cid::                         Content-ID.
* about::
* ldap::                        Lightweight Directory Access Protocol
* imap::                        IMAP mailboxes.
* man::                         Unix man pages.
@end menu

@node http/https
@section @code{http} and @code{https}

The scheme @code{http} is Hypertext Transfer Protocol.  The library
supports version 1.1, specified in RFC 2616.  (This supersedes 1.0,
defined in RFC 1945) HTTP URLs have the following form, where most of
the parts are optional:
@example
http://@var{user}:@var{password}@@@var{host}:@var{port}/@var{path}?@var{searchpart}#@var{fragment}
@end example
@c The @code{:@var{port}} part is optional, and @var{port} defaults to
@c 80.  The @code{/@var{path}} part, if present, is a slash-separated
@c series elements.  The @code{?@var{searchpart}}, if present, is the
@c query for a search or the content of a form submission.  The
@c @code{#fragment} part, if present, is a location in the document.

The scheme @code{https} is a secure version of @code{http}, with
transmission via SSL.  It is defined in RFC 2069.  Its default port is
443.  This scheme depends on SSL support in Emacs via the
@file{ssl.el} library and is actually implemented by forcing the
@code{ssl} gateway method to be used.  @xref{Gateways in general}.

@defopt url-honor-refresh-requests
This controls honouring of HTTP @samp{Refresh} headers by which
servers can direct clients to reload documents from the same URL or a
or different one.  @code{nil} means they will not be honoured,
@code{t} (the default) means they will always be honoured, and
otherwise the user will be asked on each request.
@end defopt


@menu
* Cookies::
* HTTP language/coding::
* HTTP URL Options::
* Dealing with HTTP documents::
@end menu

@node Cookies
@subsection Cookies

@defopt url-cookie-file
The file in which cookies are stored, defaulting to @file{cookies} in
the directory specified by @code{url-configuration-directory}.
@end defopt

@defopt url-cookie-confirmation
Specifies whether confirmation is require to accept cookies.
@end defopt

@defopt url-cookie-multiple-line
Specifies whether to put all cookies for the server on one line in the
HTTP request to satisfy broken servers like
@url{http://www.hotmail.com}.
@end defopt

@defopt url-cookie-trusted-urls
A list of regular expressions matching URLs from which to accept
cookies always.
@end defopt

@defopt url-cookie-untrusted-urls
A list of regular expressions matching URLs from which to reject
cookies always.
@end defopt

@defopt url-cookie-save-interval
The number of seconds between automatic saves of cookies to disk.
Default is one hour.
@end defopt


@node HTTP language/coding
@subsection Language and Encoding Preferences

HTTP allows clients to express preferences for the language and
encoding of documents which servers may honour.  For each of these
variables, the value is a string; it can specify a single choice, or
it can be a comma-separated list.

Normally this list ordered by descending preference.  However, each
element can be followed by @samp{;q=@var{priority}} to specify its
preference level, a decimal number from 0 to 1; e.g., for
@code{url-mime-language-string}, @w{@code{"de, en-gb;q=0.8,
en;q=0.7"}}.  An element that has no @samp{;q} specification has
preference level 1.

@defopt url-mime-charset-string
@cindex character sets
@cindex coding systems
This variable specifies a preference for character sets when documents
can be served in more than one encoding.

HTTP allows specifying a series of MIME charsets which indicate your
preferred character set encodings, e.g., Latin-9 or Big5, and these
can be weighted.  The default series is generated automatically from
the associated MIME types of all defined coding systems, sorted by the
coding system priority specified in Emacs.  @xref{Recognize Coding, ,
Recognizing Coding Systems, emacs, The GNU Emacs Manual}.
@end defopt

@defopt url-mime-language-string
@cindex language preferences
A string specifying the preferred language when servers can serve
files in several languages.  Use RFC 1766 abbreviations, e.g.,
@samp{en} for English, @samp{de} for German.

The string can be @code{"*"} to get the first available language (as
opposed to the default).
@end defopt

@node HTTP URL Options
@subsection HTTP URL Options

HTTP supports an @samp{OPTIONS} method describing things supported by
the URL@.

@defun url-http-options url
Returns a property list describing options available for URL.  The
property list members are:

@table @code
@item methods
A list of symbols specifying what HTTP methods the resource
supports.

@item dav
@cindex DAV
A list of numbers specifying what DAV protocol/schema versions are
supported.

@item dasl
@cindex DASL
A list of supported DASL search types supported (string form).

@item ranges
A list of the units available for use in partial document fetches.

@item p3p
@cindex P3P
The @dfn{Platform For Privacy Protection} description for the resource.
Currently this is just the raw header contents.
@end table

@end defun

@node Dealing with HTTP documents
@subsection Dealing with HTTP documents

HTTP URLs are retrieved into a buffer containing the HTTP headers
followed by the body.  Since the headers are quasi-MIME, they may be
processed using the MIME library.  @xref{Top,, Emacs MIME,
emacs-mime, The Emacs MIME Manual}.  The URL package provides a
function to do this in general:

@defun url-decode-text-part handle &optional coding
This function decodes charset-encoded text in the current buffer.  In
Emacs, the buffer is expected to be unibyte initially and is set to
multibyte after decoding.
HANDLE is the MIME handle of the original part.  CODING is an explicit
coding to use, overriding what the MIME headers specify.
The coding system used for the decoding is returned.

Note that this function doesn't deal with @samp{http-equiv} charset
specifications in HTML @samp{<meta>} elements.
@end defun

@node file/ftp
@section file and ftp
@cindex files
@cindex FTP
@cindex File Transfer Protocol
@cindex compressed files
@cindex dired

@example
ftp://@var{user}:@var{password}@@@var{host}:@var{port}/@var{file}
file://@var{user}:@var{password}@@@var{host}:@var{port}/@var{file}
@end example

These schemes are defined in RFC 1808.
@samp{ftp:} and @samp{file:} are synonymous in this library.  They
allow reading arbitrary files from hosts.  Either @samp{ange-ftp}
(Emacs) or @samp{efs} (XEmacs) is used to retrieve them from remote
hosts.  Local files are accessed directly.

Compressed files are handled, but support is hard-coded so that
@code{jka-compr-compression-info-list} and so on have no affect.
Suffixes recognized are @samp{.z}, @samp{.gz}, @samp{.Z} and
@samp{.bz2}.

@defopt url-directory-index-file
The filename to look for when indexing a directory, default
@samp{"index.html"}.  If this file exists, and is readable, then it
will be viewed instead of using @code{dired} to view the directory.
@end defopt

@node info
@section info
@cindex Info
@cindex Texinfo
@findex Info-goto-node

@example
info:@var{file}#@var{node}
@end example

Info URLs are not officially defined.  They invoke
@code{Info-goto-node} with argument @samp{(@var{file})@var{node}}.
@samp{#@var{node}} is optional, defaulting to @samp{Top}.

@node mailto
@section mailto

@cindex mailto
@cindex email
A mailto URL will send an email message to the address in the
URL, for example @samp{mailto:foo@@bar.com} would compose a
message to @samp{foo@@bar.com}.

@defopt url-mail-command
@vindex mail-user-agent
The function called whenever url needs to send mail.  This should
normally be left to default from @var{mail-user-agent}.  @xref{Mail
Methods, , Mail-Composition Methods, emacs, The GNU Emacs Manual}.
@end defopt

An @samp{X-Url-From} header field containing the URL of the document
that contained the mailto URL is added if that URL is known.

RFC 2368 extends the definition of mailto URLs in RFC 1738.
The form of a mailto URL is
@example
@samp{mailto:@var{mailbox}[?@var{header}=@var{contents}[&@var{header}=@var{contents}]]}
@end example
@noindent where an arbitrary number of @var{header}s can be added.  If the
@var{header} is @samp{body}, then @var{contents} is put in the body
otherwise a @var{header} header field is created with @var{contents}
as its contents.  Note that the URL library does not consider any
headers `dangerous' so you should check them before sending the
message.

@c Fixme: update
Email messages are defined in @sc{rfc}822.

@node news/nntp/snews
@section @code{news}, @code{nntp} and @code{snews}
@cindex news
@cindex network news
@cindex usenet
@cindex NNTP
@cindex snews

@c draft-gilman-news-url-01
The network news URL scheme take the following forms following RFC
1738 except that for compatibility with other clients, host and port
fields may be included in news URLs though they are properly only
allowed for nntp an snews.

@table @samp
@item news:@var{newsgroup}
Retrieves a list of messages in @var{newsgroup};
@item news:@var{message-id}
Retrieves the message with the given @var{message-id};
@item news:*
Retrieves a list of all available newsgroups;
@item nntp://@var{host}:@var{port}/@var{newsgroup}
@itemx nntp://@var{host}:@var{port}/@var{message-id}
@itemx nntp://@var{host}:@var{port}/*
Similar to the @samp{news} versions.
@end table

@samp{:@var{port}} is optional and defaults to :119.

@samp{snews} is the same as @samp{nntp} except that the default port
is :563.
@cindex SSL
(It is tunneled through SSL.)

An @samp{nntp} URL is the same as a news URL, except that the URL may
specify an article by its number.

@defopt url-news-server
This variable can be used to override the default news server.
Usually this will be set by the Gnus package, which is used to fetch
news.
@cindex environment variable
@vindex NNTPSERVER
It may be set from the conventional environment variable
@code{NNTPSERVER}.
@end defopt

@node rlogin/telnet/tn3270
@section rlogin, telnet and tn3270
@cindex rlogin
@cindex telnet
@cindex tn3270
@cindex terminal emulation
@findex terminal-emulator

These URL schemes from RFC 1738 for logon via a terminal emulator have
the form
@example
telnet://@var{user}:@var{password}@@@var{host}:@var{port}
@end example
but the @code{:@var{password}} component is ignored.

To handle rlogin, telnet and tn3270 URLs, a @code{rlogin},
@code{telnet} or @code{tn3270} (the program names and arguments are
hardcoded) session is run in a @code{terminal-emulator} buffer.
Well-known ports are used if the URL does not specify a port.

@node irc
@section irc
@cindex IRC
@cindex Internet Relay Chat
@cindex ZEN IRC
@cindex ERC
@cindex rcirc
@c Fixme: reference (was http://www.w3.org/Addressing/draft-mirashi-url-irc-01.txt)
@dfn{Internet Relay Chat} (IRC) is handled by handing off the @sc{irc}
session to a function named in @code{url-irc-function}.

@defopt url-irc-function
A function to actually open an IRC connection.
This function
must take five arguments, @var{host}, @var{port}, @var{channel},
@var{user} and @var{password}.  The @var{channel} argument specifies the
channel to join immediately, this can be @code{nil}.  By default this is
@code{url-irc-rcirc}.
@end defopt
@defun url-irc-rcirc host port channel user password
Processes the arguments and lets @code{rcirc} handle the session.
@end defun
@defun url-irc-erc host port channel user password
Processes the arguments and lets @code{ERC} handle the session.
@end defun
@defun url-irc-zenirc host port channel user password
Processes the arguments and lets @code{zenirc} handle the session.
@end defun

@node data
@section data
@cindex data URLs

@example
data:@r{[}@var{media-type}@r{]}@r{[};@var{base64}@r{]},@var{data}
@end example

Data URLs contain MIME data in the URL itself.  They are defined in
RFC 2397.

@var{media-type} is a MIME @samp{Content-Type} string, possibly
including parameters.  It defaults to
@samp{text/plain;charset=US-ASCII}.  The @samp{text/plain} can be
omitted but the charset parameter supplied.  If @samp{;base64} is
present, the @var{data} are base64-encoded.

@node nfs
@section nfs
@cindex NFS
@cindex Network File System
@cindex automounter

@example
nfs://@var{user}:@var{password}@@@var{host}:@var{port}/@var{file}
@end example

The @samp{nfs:} scheme is defined in RFC 2224.  It is similar to
@samp{ftp:} except that it points to a file on a remote host that is
handled by the automounter on the local host.

@defvar url-nfs-automounter-directory-spec
@end defvar
A string saying how to invoke the NFS automounter.  Certain @samp{%}
sequences are recognized:

@table @samp
@item %h
The hostname of the NFS server;
@item %n
The port number of the NFS server;
@item %u
The username to use to authenticate;
@item %p
The password to use to authenticate;
@item %f
The filename on the remote server;
@item %%
A literal @samp{%}.
@end table

Each can be used any number of times.

@node cid
@section cid
@cindex Content-ID

RFC 2111

@node about
@section about

@node ldap
@section ldap
@cindex LDAP
@cindex Lightweight Directory Access Protocol

The LDAP scheme is defined in RFC 2255.

@node imap
@section imap
@cindex IMAP

RFC 2192

@node man
@section man
@cindex @command{man}
@cindex Unix man pages
@findex man

@example
@samp{man:@var{page-spec}}
@end example

This is a non-standard scheme.  @var{page-spec} is passed directly to
the Lisp @code{man} function.

@node Defining New URLs
@chapter Defining New URLs

@menu
* Naming conventions::
* Required functions::
* Optional functions::
* Asynchronous fetching::
* Supporting file-name-handlers::
@end menu

@node Naming conventions
@section Naming conventions

@node Required functions
@section Required functions

@node Optional functions
@section Optional functions

@node Asynchronous fetching
@section Asynchronous fetching

@node Supporting file-name-handlers
@section Supporting file-name-handlers

@node General Facilities
@chapter General Facilities

@menu
* Disk Caching::
* Proxies::
* Gateways in general::
* History::
@end menu

@node Disk Caching
@section Disk Caching
@cindex Caching
@cindex Persistent Cache
@cindex Disk Cache

The disk cache stores retrieved documents locally, whence they can be
retrieved more quickly.  When requesting a URL that is in the cache,
the library checks to see if the page has changed since it was last
retrieved from the remote machine.  If not, the local copy is used,
saving the transmission over the network.
@cindex Cleaning the cache
@cindex Clearing the cache
@cindex Cache cleaning
Currently the cache isn't cleared automatically.
@c Running the @code{clean-cache} shell script
@c fist is recommended, to allow for future cleaning of the cache.  This
@c shell script will remove all files that have not been accessed since it
@c was last run.  To keep the cache pared down, it is recommended that this
@c script be run from @i{at} or @i{cron} (see the manual pages for
@c crontab(5) or at(1) for more information)

@defopt url-automatic-caching
Setting this variable non-@code{nil} causes documents to be cached
automatically.
@end defopt

@defopt url-cache-directory
This variable specifies the
directory to store the cache files.  It defaults to sub-directory
@file{cache} of @code{url-configuration-directory}.
@end defopt

@c Fixme: function v. option, but neither used.
@c @findex url-cache-expired
@c @defopt url-cache-expired
@c This is a function to decide whether or not a cache entry has expired.
@c It takes two times as it parameters and returns non-@code{nil} if the
@c second time is ``too old'' when compared with the first time.
@c @end defopt

@defopt url-cache-creation-function
The cache relies on a scheme for mapping URLs to files in the cache.
This variable names a function which sets the type of cache to use.
It takes a URL as argument and returns the absolute file name of the
corresponding cache file.  The two supplied possibilities are
@code{url-cache-create-filename-using-md5} and
@code{url-cache-create-filename-human-readable}.
@end defopt

@defun url-cache-create-filename-using-md5 url
Creates a cache file name from @var{url} using MD5 hashing.
This is creates entries with very few cache collisions and is fast.
@cindex MD5
@smallexample
(url-cache-create-filename-using-md5 "http://www.example.com/foo/bar")
  @result{} "/home/fx/.url/cache/fx/http/com/example/www/b8a35774ad20db71c7c3409a5410e74f"
@end smallexample
@end defun

@defun url-cache-create-filename-human-readable url
Creates a cache file name from @var{url} more obviously connected to
@var{url} than for @code{url-cache-create-filename-using-md5}, but
more likely to conflict with other files.
@smallexample
(url-cache-create-filename-human-readable "http://www.example.com/foo/bar")
  @result{} "/home/fx/.url/cache/fx/http/com/example/www/foo/bar"
@end smallexample
@end defun

@c Fixme: never actually used currently?
@c @defopt url-standalone-mode
@c @cindex Relying on cache
@c @cindex Cache only mode
@c @cindex Standalone mode
@c If this variable is non-@code{nil}, the library relies solely on the
@c cache for fetching documents and avoids checking if they have changed
@c on remote servers.
@c @end defopt

@c With a large cache of documents on the local disk, it can be very handy
@c when traveling, or any other time the network connection is not active
@c (a laptop with a dial-on-demand PPP connection, etc).  Emacs/W3 can rely
@c solely on its cache, and avoid checking to see if the page has changed
@c on the remote server.  In the case of a dial-on-demand PPP connection,
@c this will keep the phone line free as long as possible, only bringing up
@c the PPP connection when asking for a page that is not located in the
@c cache.  This is very useful for demonstrations as well.

@node Proxies
@section Proxies and Gatewaying

@c fixme: check/document url-ns stuff
@cindex proxy servers
@cindex proxies
@cindex environment variables
@vindex HTTP_PROXY
Proxy servers are commonly used to provide gateways through firewalls
or as caches serving some more-or-less local network.  Each protocol
(HTTP, FTP, etc.)@: can have a different gateway server.  Proxying is
conventionally configured commonly amongst different programs through
environment variables of the form @code{@var{protocol}_proxy}, where
@var{protocol} is one of the supported network protocols (@code{http},
@code{ftp} etc.).  The library recognizes such variables in either
upper or lower case.  Their values are of one of the forms:
@itemize @bullet
@item @code{@var{host}:@var{port}}
@item A full URL;
@item Simply a host name.
@end itemize

@vindex NO_PROXY
The @code{NO_PROXY} environment variable specifies URLs that should be
excluded from proxying (on servers that should be contacted directly).
This should be a comma-separated list of hostnames, domain names, or a
mixture of both.  Asterisks can be used as wildcards, but other
clients may not support that.  Domain names may be indicated by a
leading dot.  For example:
@example
NO_PROXY="*.aventail.com,home.com,.seanet.com"
@end example
@noindent says to contact all machines in the @samp{aventail.com} and
@samp{seanet.com} domains directly, as well as the machine named
@samp{home.com}.  If @code{NO_PROXY} isn't defined, @code{no_PROXY}
and @code{no_proxy} are also tried, in that order.

Proxies may also be specified directly in Lisp.

@defopt url-proxy-services
This variable is an alist of URL schemes and proxy servers that
gateway them.  The items are of the form @w{@code{(@var{scheme}
. @var{host}:@var{portnumber})}}, says that the URL @var{scheme} is
gatewayed through @var{portnumber} on the specified @var{host}.  An
exception is the pseudo scheme @code{"no_proxy"}, which is paired with
a regexp matching host names not to be proxied.  This variable is
initialized from the environment as above.

@example
(setq url-proxy-services
      '(("http"     . "proxy.aventail.com:80")
        ("no_proxy" . "^.*\\(aventail\\|seanet\\)\\.com")))
@end example
@end defopt

@node Gateways in general
@section Gateways in General
@cindex gateways
@cindex firewalls

The library provides a general gateway layer through which all
networking passes.  It can both control access to the network and
provide access through gateways in firewalls.  This may make direct
connections in some cases and pass through some sort of gateway in
others.@footnote{Proxies (which only operate over HTTP) are
implemented using this.}  The library's basic function responsible for
making connections is @code{url-open-stream}.

@defun url-open-stream name buffer host service
@cindex opening a stream
@cindex stream, opening
Open a stream to @var{host}, possibly via a gateway.  The other
arguments are as for @code{open-network-stream}.  This will not make a
connection if @code{url-gateway-unplugged} is non-@code{nil}.
@end defun

@defvar url-gateway-local-host-regexp
This is a regular expression that matches local hosts that do not
require the use of a gateway.  If @code{nil}, all connections are made
through the gateway.
@end defvar

@defvar url-gateway-method
This variable controls which gateway method is used.  It may be useful
to bind it temporarily in some applications.  It has values taken from
a list of symbols.  Possible values are:

@table @code
@item telnet
@cindex @command{telnet}
Use this method if you must first telnet and log into a gateway host,
and then run telnet from that host to connect to outside machines.

@item rlogin
@cindex @command{rlogin}
This method is identical to @code{telnet}, but uses @command{rlogin}
to log into the remote machine without having to send the username and
password over the wire every time.

@item socks
@cindex @sc{socks}
Use if the firewall has a @sc{socks} gateway running on it.  The
@sc{socks} v5 protocol is defined in RFC 1928.

@c @item ssl
@c This probably shouldn't be documented
@c Fixme: why not? -- fx

@item native
This method uses Emacs's builtin networking directly.  This is the
default.  It can be used only if there is no firewall blocking access.
@end table
@end defvar

The following variables control the gateway methods.

@defopt url-gateway-telnet-host
The gateway host to telnet to.  Once logged in there, you then telnet
out to the hosts you want to connect to.
@end defopt
@defopt url-gateway-telnet-parameters
This should be a list of parameters to pass to the @command{telnet} program.
@end defopt
@defopt url-gateway-telnet-password-prompt
This is a regular expression that matches the password prompt when
logging in.
@end defopt
@defopt url-gateway-telnet-login-prompt
This is a regular expression that matches the username prompt when
logging in.
@end defopt
@defopt url-gateway-telnet-user-name
The username to log in with.
@end defopt
@defopt url-gateway-telnet-password
The password to send when logging in.
@end defopt
@defopt url-gateway-prompt-pattern
This is a regular expression that matches the shell prompt.
@end defopt

@defopt url-gateway-rlogin-host
Host to @samp{rlogin} to before telnetting out.
@end defopt
@defopt url-gateway-rlogin-parameters
Parameters to pass to @samp{rsh}.
@end defopt
@defopt url-gateway-rlogin-user-name
User name to use when logging in to the gateway.
@end defopt
@defopt url-gateway-prompt-pattern
This is a regular expression that matches the shell prompt.
@end defopt

@defopt socks-server
This specifies the default server, it takes the form
@w{@code{("Default server" @var{server} @var{port} @var{version})}}
where @var{version} can be either 4 or 5.
@end defopt
@defvar socks-password
If this is @code{nil} then you will be asked for the password,
otherwise it will be used as the password for authenticating you to
the @sc{socks} server.
@end defvar
@defvar socks-username
This is the username to use when authenticating yourself to the
@sc{socks} server.  By default this is your login name.
@end defvar
@defvar socks-timeout
This controls how long, in seconds, to wait for responses from the
@sc{socks} server; it is 5 by default.
@end defvar
@c fixme: these have been effectively commented-out in the code
@c @defopt socks-server-aliases
@c This a list of server aliases.  It is a list of aliases of the form
@c @var{(alias hostname port version)}.
@c @end defopt
@c @defopt socks-network-aliases
@c This a list of network aliases.  Each entry in the list takes the form
@c @var{(alias (network))} where @var{alias} is a string that names the
@c @var{network}.  The networks can contain a pair (not a dotted pair) of
@c @sc{ip} addresses which specify a range of @sc{ip} addresses, an @sc{ip}
@c address and a netmask, a domain name or a unique hostname or @sc{ip}
@c address.
@c @end defopt
@c @defopt socks-redirection-rules
@c This a list of redirection rules.  Each rule take the form
@c @var{(Destination network Connection type)} where @var{Destination
@c network} is a network alias from @code{socks-network-aliases} and
@c @var{Connection type} can be @code{nil} in which case a direct
@c connection is used, or it can be an alias from
@c @code{socks-server-aliases} in which case that server is used as a
@c proxy.
@c @end defopt
@defopt socks-nslookup-program
@cindex @command{nslookup}
This the @samp{nslookup} program.  It is @code{"nslookup"} by default.
@end defopt

@menu
* Suppressing network connections::
@end menu
@c * Broken hostname resolution::

@node Suppressing network connections
@subsection Suppressing Network Connections

@cindex network connections, suppressing
@cindex suppressing network connections
@cindex bugs, HTML
@cindex HTML `bugs'
In some circumstances it is desirable to suppress making network
connections.  A typical case is when rendering HTML in a mail user
agent, when external URLs should not be activated, particularly to
avoid `bugs' which `call home' by fetch single-pixel images and the
like.  To arrange this, bind the following variable for the duration
of such processing.

@defvar url-gateway-unplugged
If this variable is non-@code{nil} new network connections are never
opened by the URL library.
@end defvar

@c @node Broken hostname resolution
@c @subsection Broken Hostname Resolution

@c @cindex hostname resolver
@c @cindex resolver, hostname
@c Some C libraries do not include the hostname resolver routines in
@c their static libraries.  If Emacs was linked statically, and was not
@c linked with the resolver libraries, it will not be able to get to any
@c machines off the local network.  This is characterized by being able
@c to reach someplace with a raw ip number, but not its hostname
@c (@url{http://129.79.254.191/} works, but
@c @url{http://www.cs.indiana.edu/} doesn't).  This used to happen on
@c SunOS4 and Ultrix, but is now probably now rare.  If Emacs can't be
@c rebuilt linked against the resolver library, it can use the external
@c @command{nslookup} program instead.

@c @defopt url-gateway-broken-resolution
@c @cindex @code{nslookup} program
@c @cindex program, @code{nslookup}
@c If non-@code{nil}, this variable says to use the program specified by
@c @code{url-gateway-nslookup-program} program to do hostname resolution.
@c @end defopt

@c @defopt url-gateway-nslookup-program
@c The name of the program to do hostname lookup if Emacs can't do it
@c directly.  This program should expect a single argument on the command
@c line---the hostname to resolve---and should produce output similar to
@c the standard Unix @command{nslookup} program:
@c @example
@c Name: www.cs.indiana.edu
@c Address: 129.79.254.191
@c @end example
@c @end defopt

@node History
@section History

@findex url-do-setup
The library can maintain a global history list tracking URLs accessed.
URL completion can be done from it.  The history mechanism is set up
automatically via @code{url-do-setup} when it is configured to be on.
Note that the size of the history list is currently not limited.

@vindex url-history-hash-table
The history `list' is actually a hash table,
@code{url-history-hash-table}.  It contains access times keyed by URL
strings.  The times are in the format returned by @code{current-time}.

@defun url-history-update-url url time
This function updates the history table with an entry for @var{url}
accessed at the given @var{time}.
@end defun

@defopt url-history-track
If non-@code{nil}, the library will keep track of all the URLs
accessed.  If it is @code{t}, the list is saved to disk at the end of
each Emacs session.  The default is @code{nil}.
@end defopt

@defopt url-history-file
The file storing the history list between sessions.  It defaults to
@file{history} in @code{url-configuration-directory}.
@end defopt

@defopt url-history-save-interval
@findex url-history-setup-save-timer
The number of seconds between automatic saves of the history list.
Default is one hour.  Note that if you change this variable directly,
rather than using Custom, after @code{url-do-setup} has been run, you
need to run the function @code{url-history-setup-save-timer}.
@end defopt

@defun url-history-parse-history &optional fname
Parses the history file @var{fname} (default @code{url-history-file})
and sets up the history list.
@end defun

@defun url-history-save-history &optional fname
Saves the current history to file @var{fname} (default
@code{url-history-file}).
@end defun

@defun url-completion-function string predicate function
You can use this function to do completion of URLs from the history.
@end defun

@node Customization
@chapter Customization

@section Environment Variables

@cindex environment variables
The following environment variables affect the library's operation at
startup.

@table @code
@item TMPDIR
@vindex TMPDIR
@vindex url-temporary-directory
If this is defined, @var{url-temporary-directory} is initialized from
it.
@end table

@section General User Options

The following user options, settable with Customize, affect the
general operation of the package.

@defopt url-debug
@cindex debugging
Specifies the types of debug messages the library which are logged to
the @code{*URL-DEBUG*} buffer.
@code{t} means log all messages.
A number means log all messages and show them with @code{message}.
If may also be a list of the types of messages to be logged.
@end defopt
@defopt url-personal-mail-address
@end defopt
@defopt url-privacy-level
@end defopt
@defopt url-uncompressor-alist
@end defopt
@defopt url-passwd-entry-func
@end defopt
@defopt url-standalone-mode
@end defopt
@defopt url-bad-port-list
@end defopt
@defopt url-max-password-attempts
@end defopt
@defopt url-temporary-directory
@end defopt
@defopt url-show-status
@end defopt
@defopt url-confirmation-func
The function to use for asking yes or no functions.  This is normally
either @code{y-or-n-p} or @code{yes-or-no-p}, but could be another
function taking a single argument (the prompt) and returning @code{t}
only if an affirmative answer is given.
@end defopt
@defopt url-gateway-method
@c fixme: describe gatewaying
A symbol specifying the type of gateway support to use for connections
from the local machine.  The supported methods are:

@table @code
@item telnet
Run telnet in a subprocess to connect;
@item rlogin
Rlogin to another machine to connect;
@item socks
Connect through a socks server;
@item ssl
Connect with SSL;
@item native
Connect directly.
@end table
@end defopt

@node GNU Free Documentation License
@appendix GNU Free Documentation License
@include doclicense.texi

@node Function Index
@unnumbered Command and Function Index
@printindex fn

@node Variable Index
@unnumbered Variable Index
@printindex vr

@node Concept Index
@unnumbered Concept Index
@printindex cp

@setchapternewpage odd
@contents
@bye

@ignore
   arch-tag: c96be356-7e2d-4196-bcda-b13246c5c3f0
@end ignore