Redirex

Redirect HTTP Requests to New Web Server

Version 1.1

Version 2.0 of Redirex was released in July of 2004. It is compatible with current versions of Perl in "strict" syntax mode, and is easier to configure. Version 1.1 should be used only if you require compatibility with versions of Perl prior to 5.6.

Introduction

There are few better ways to appreciate the breadth of the Web and the depth of its content as to move a Web server from one machine to another with a different host name and IP address. Whatever the reason motivating the change, you'll quickly discover that requests continue to rain onto the obsolete server, diminishing ever so slowly only as other webmasters discover and correct broken links to your site, and search engines eventually discover the change and re-index your site at its new location. Redirex is a utility which can smooth the transition of a Web site from one server to another, both for users of the site and its administrators. Redirex is a Perl program which receives requests at the address of the old server and responds with HTTP "redirect" (301 status code) replies indicating the new server's name.

Installation and Configuration

To install and configure Redirex, perform the following steps.

  1. Download the Redirex source archive and extract its contents into a new directory. The archive contains redirex, the main Perl script, redirex.conf, its configuration file, and a copy of the standard Perl library subroutine getopt.pl. The latter is included solely to simplify use on systems without a complete installation of Perl; if your system has the Perl library properly installed, you may delete this file from the Redirex directory.

  2. Edit redirex.conf and set the configuration variables appropriately for your application. The configuration variables are discussed in detail below.

  3. (Optional) Test Redirex by running it on a user port (for example, 9080), then accessing it from your Web browser with a URL like: "http://my.hostname.net:9080/whatever.html". Verify that the request was redirected as specified by the configuration file.

  4. Install Redirex on the old server, configured to listen to the port (usually 80) on which that server used to accept HTTP requests. If the old server no longer exists, configure a different machine to respond to the old server's IP address, perhaps using "IP aliases" or "logical interfaces" to make an existing machine respond to the old server's address. This is discussed in more detail below in the "Virtual Host Setup" section.

  5. Verify that Redirex is correctly redirecting requests to the old server and writing log items in the configured file.

Command Line Options

Usually Redirex is run with no options on the command line; in this case it obtains all its configuration parameters from the default configuration file, redirex.conf. You can specify a different configuration file or cause Redirex to listen for HTTP requests on a different port by specifying the following options.

-c conffile
Redirex will read its configuration parameters from the specified conffile instead of the default, redirex.conf. This is handy if you wish to run several copies of Redirex to field requests for multiple servers; each copy can be started with a configuration file corresponding to the server it is to redirect.

-p port
Redirex will listen on the specified port instead of the port given in the configuration file. This permits testing Redirex with ports above 1024 which don't require root (superuser) privilege without the need to edit the configuration file.

Configuration File Variables

The following variables are defined in the configuration file. By default, the configuration is read from redirex.conf, but you can specify a different configuration file with the -c option on the command line. The configuration file is a Perl program in its own right; specifications in the file must conform to valid Perl syntax.
$AF_INET . . . $EPROTO
If your system doesn't use standard System V values for these variables (as is the case for some BSD-derived systems such as SunOS 4.x), replace them with the correct values from your /usr/include/sys/socket.h and /usr/include/sys/errno.h files. The proper Perl way to do is to include .ph files created by h2ph from the corresponding C include files, but I've found many Perl installations have not set up these files correctly. Hard coding is ugly, but it avoids having to fight with the Perl installation just to get Redirex running. The comments on each variable name the directory in C notation, usually below /usr/include, where the definition can usually be found.

$defaultport
This specifies the port number on which to listen for HTTP requests. The standard port for HTTP requests is 80. Note that to bind to any port with a number less than 1024, Redirex must be run with root (super-user) privilege; for testing use a port accessible to regular users (for example 9080), and specify that port in URLs to exercise Redirex. After you're sure everything is working correctly, change the port number to the standard 80 and run Redirex as super-user. If the -p command line option is specified, this value will not be used.

$IPlisten
This variable specifies which IP address to listen to, in dotted decimal notation, for example '192.168.37.252'. If you wish to listen to all IP addresses, set this to '0.0.0.0'. Note that if you specify a value of '0.0.0.0', Redirex will intercept all requests on the configured port for all IP addresses on which the system running it listens. If the machine running Redirex also acts as a Web server, you'll have to specify the explicit IP address on which Redirex will listen. If you need to run two or more copies of Redirex on a machine which is also a Web server, you'll need to prepare a separate configuration file for each server you're redirecting and start one copy Redirex for each, specifying the configuration file with the -c command line option.

$newServer
This variable gives the URL prefix consisting of the protocol, host name, and (optionally) port number of the new server to which requests are being redirected. Redirex assumes the new server has the same content (or a superset thereof) and directory structure as the server it is redirecting; the file name portion of the URL is simply appended to the $newServer string. For example, suppose Redirex is intercepting requests to port 80 at IP address 192.168.133.12, host name gnarly.oldsite.net and $newServer is set to 'spiffy.newsite.net'. Then the URL request "http://gnarly.oldsite.net/info/catalogue.html" will be redirected to "http://spiffy.newsite.net/info/catalogue.html".

$newHomePage and $newHomePageDescription
HTTP redirection should be transparent to the user, simply replacing the requested URL with the new destination. Since some browsers may not correctly support redirection, the HTTP standard recommends that redirection messages sent in response to GET and POST requests (but not HEAD) include descriptive text and a hyperlink to the new destination URL. That way, if the browser fails to perform the redirection, the user can simply click on the link to access the new location. When generating this message, Redirex includes a link to the URL specified by $newHomePage with descriptive text $newHomePageDescription. This should point to the home page of the new destination server. If the redirection should fail due to changes in the directory structure from that of the old server, the user can use this link to find the home page of the new server.

$logfile
Redirex logs all requests it redirects in the named $logfile, which should be specified as an absolute path name. Log items are appended to the file, written in the "Common logfile format" used by most present-day Web servers. To avoid many time-consuming domain name lookups, numeric IP addresses are used in the log file instead of host names. If you require host names, process the Redirex file with the logresolve program included with the Apache HTTP Server or the Logtail program available at this site. Redirected requests show the redirection status code of 301. The length of the reply is always 512, a reasonable approximation of the length of the redirection document.

$DOredirect
If $DOredirect is set to 1, requests will be hard-redirected with a 301 status code. If zero, the reply will be a normal 200 status document which informs the user of the redirection but doesn't request the browser to automatically divert there. Requests processed with $DOredirect set to 0 appear in the log file with a status of 200.

Virtual Host Setup

If you can't afford to dedicate a computer to impersonate the server you're redirecting from, you'll need to configure an existing host on your network to respond to the old server's IP address(es). (Dedicating a computer to running Redirex is not necessarily absurd; Redirex requires minimal system resources, and many sites have one or more retired PCs which are perfectly adequate to run Redirex under a system such as Linux.)

Traditionally, Unix systems associated a unique IP address with each network interface; the only way to cause a host to listen to multiple IP addresses was to equip it with as many hardware interfaces as addresses. Obviously, in cases where a single server hosts a large number of Web sites, this requirement was untenable--a large Internet Service Provider might require thousands of network interfaces to support all of their customers' IP addresses. Most modern Unix systems provide a mechanism which permits receiving packets for multiple IP addresses on a single hardware interface. Unfortunately, each different version of Unix seems to have invented its own mechanism for accomplishing this. To figure out how to configure virtual hosts on your server, a good place to start is the system's manual page for the ifconfig command.

Once you've managed to configure your server to listen on the the old server's IP address, you need to make sure there's no conflict between Redirex and any Web server running on the same machine. Many Web servers, including Apache, listen by default to Port 80 on all IP addresses received by the host. Thus, if you start the Web server first, Redirex will fail because the address it wishes to listen on has already been assigned to the Web server. To get around this problem, you must configure the Web server to respond only to the IP address(es) it is intended to serve (use "Listen" directives with Apache), then configure Redirex with the $IPlisten variable set to the address it is to redirect. If you need to redirect more than one IP address, run a separate copy of Redirex for each, using the -c command line option to specify separate configuration files for each server.

Download redirex.tar.gz (gzipped TAR archive)

Copying and Support Information

This software is in the public domain. Permission to use, copy, modify, and distribute this software and its documentation for any purpose and without fee is hereby granted, without any conditions or restrictions. This software is provided "as is" without express or implied warranty.

Absolutely no support or assistance of any kind whatsoever is available for Redirex--you are entirely on your own. As should be evident from reading this document, while Redirex is a small, simple program, the many varieties of Unix in use on Web servers makes it impossible to provide a cookbook installation procedure. Getting Redirex to work requires system administration skills comparable to those needed to install a Web server; if you don't understand terms such as "IP address", "DNS", "ifconfig", "netstat", "Perl", etc. you probably won't get very far with Redirex.

Bugs, Features, and Gotchas

Redirex is a Perl script. In order to use it, you need Perl version 4, patch level 36 or higher (often referred to as Perl 4.036). Redirex should also work with Perl 5, but I have not extensively tested it with that version.

Redirex has been tested only under various flavours of the Unix operating system. In order for it to work on other systems (for example, OS/2 or Windows NT), the implementation of Perl used to run it must include Unix-compatible networking, fork, and signal facilities.

Redirex assumes the directory structure and file names on the new server are identical to those on the machine being redirected. If you need more complex file name rewriting, you're going to have to add it yourself. (For complicated redirection, you're probably better off installing the Apache server, which includes extensive URL transformation and redirection support. Redirex was written to permit simple redirection without the need to install such a relatively large and complicated package.)

If you're running Redirex on a machine with limited disc space or a stripped-down configuration (for example, a firewall host), note that you don't have to fully install Perl on the machine in order to run Redirex. It's sufficient to copy an executable of the Perl interpreter into the directory with Redirex and start it with "./perl redirex".

Transfer lengths written in the log file do not represent the actual byte count; they're always 512.

The only way to cause Redirex to transfer its log to a new file is to kill Redirex, rename the log file, and then restart the program.

Acknowledgements

Redirex is derived from mhttpd--a small HTTP server written in Perl by Jerry LeVan, which in turn was inspired by Bob Diertens' simple CGI "Get" server for executables and Pratap Pereira's phttpd. Redirex is a much simpler application than mhttpd, so much of the code in the original program has been deleted in creating this single-purpose redirector. Naturally, any errors and omissions in this program are entirely my own responsibility.

Fourmilab Home Page


by John Walker
March 23rd, 1998