At least for important top-level pages it is sometimes necessary to provide the optimum of browser dependent content, i.e. one has to provide a maximum version for the latest Netscape variants, a minimum version for the Lynx browsers and a average feature version for all others.

We cannot use content negotiation because the browsers do not provide their type in that form. Instead we have to act on the HTTP header "User-Agent". The following condig does the following: If the HTTP header "User-Agent" begins with "Mozilla/3", the page foo.html is rewritten to foo.NS.html and and the rewriting stops. If the browser is "Lynx" or "Mozilla" of version 1 or 2 the URL becomes foo.20.html. All other browsers receive page foo.32.html. This is done by the following ruleset:

RewriteCond %{HTTP_USER_AGENT}  ^Mozilla/3.*
RewriteRule ^foo\.html$         foo.NS.html          [L]

RewriteCond %{HTTP_USER_AGENT}  ^Lynx/.*         [OR]
RewriteCond %{HTTP_USER_AGENT}  ^Mozilla/[12].*
RewriteRule ^foo\.html$         foo.20.html          [L]

RewriteRule ^foo\.html$         foo.32.html          [L]

Assume there are nice webpages on remote hosts we want to bring into our namespace. For FTP servers we would use the mirror program which actually maintains an explicit up-to-date copy of the remote data on the local machine. For a webserver we could use the program webcopy which acts similar via HTTP. But both techniques have one major drawback: The local copy is always just as up-to-date as often we run the program. It would be much better if the mirror is not a static one we have to establish explicitly. Instead we want a dynamic mirror with data which gets updated automatically when there is need (updated data on the remote host).

To provide this feature we map the remote webpage or even the complete remote webarea to our namespace by the use of the Proxy Throughput feature (flag [P]):

RewriteEngine  on
RewriteBase    /~quux/
RewriteRule    ^hotsheet/(.*)$  http://www.tstimpreso.com/hotsheet/$1  [P]

RewriteEngine  on
RewriteBase    /~quux/
RewriteRule    ^usa-news\.html$   http://www.quux-corp.com/news/index.html  [P]

This is a tricky way of virtually running a corporate (external) Internet webserver (www.quux-corp.dom), while actually keeping and maintaining its data on a (internal) Intranet webserver (www2.quux-corp.dom) which is protected by a firewall. The trick is that on the external webserver we retrieve the requested data on-the-fly from the internal one.

First, we have to make sure that our firewall still protects the internal webserver and that only the external webserver is allowed to retrieve data from it. For a packet-filtering firewall we could for instance configure a firewall ruleset like the following:

ALLOW Host www.quux-corp.dom Port >1024 --> Host www2.quux-corp.dom Port 80
DENY  Host *                 Port *     --> Host www2.quux-corp.dom Port 80

Just adjust it to your actual configuration syntax. Now we can establish the mod_rewrite rules which request the missing data in the background through the proxy throughput feature:

RewriteRule ^/~([^/]+)/?(.*)          /home/$1/.www/$2
RewriteCond %{REQUEST_FILENAME}       !-f
RewriteCond %{REQUEST_FILENAME}       !-d
RewriteRule ^/home/([^/]+)/.www/?(.*) http://www2.quux-corp.dom/~$1/pub/$2 [P]

Suppose we want to load balance the traffic to www.foo.com over www[0-5].foo.com (a total of 6 servers). How can this be done?

There are a lot of possible solutions for this problem. We will discuss first a commonly known DNS-based variant and then the special one with mod_rewrite:

1. DNS Round-Robin

   The simplest method for load-balancing is to use the DNS round-robin feature of BIND. Here you just configure www[0-9].foo.com as usual in your DNS with A(address) records, e.g.

   www0   IN  A       1.2.3.1
   www1   IN  A       1.2.3.2
   www2   IN  A       1.2.3.3
   www3   IN  A       1.2.3.4
   www4   IN  A       1.2.3.5
   www5   IN  A       1.2.3.6
   

   Then you additionally add the following entry:

   www   IN  A       1.2.3.1
   www   IN  A       1.2.3.2
   www   IN  A       1.2.3.3
   www   IN  A       1.2.3.4
   www   IN  A       1.2.3.5
   

   Now when www.foo.com gets resolved, BIND gives out www0-www5 - but in a slightly permutated/rotated order every time. This way the clients are spread over the various servers. But notice that this not a perfect load balancing scheme, because DNS resolve information gets cached by the other nameservers on the net, so once a client has resolved www.foo.com to a particular wwwN.foo.com, all subsequent requests also go to this particular name wwwN.foo.com. But the final result is ok, because the total sum of the requests are really spread over the various webservers.

2. DNS Load-Balancing

   A sophisticated DNS-based method for load-balancing is to use the program lbnamed which can be found at http://www.stanford.edu/~schemers/docs/lbnamed/lbnamed.html. It is a Perl 5 program in conjunction with auxilliary tools which provides a real load-balancing for DNS.

3. Proxy Throughput Round-Robin

   In this variant we use mod_rewrite and its proxy throughput feature. First we dedicate www0.foo.com to be actually www.foo.com by using a single

   www    IN  CNAME   www0.foo.com.
   

   entry in the DNS. Then we convert www0.foo.com to a proxy-only server, i.e. we configure this machine so all arriving URLs are just pushed through the internal proxy to one of the 5 other servers (www1-www5). To accomplish this we first establish a ruleset which contacts a load balancing script lb.pl for all URLs.

   RewriteEngine on
   RewriteMap    lb      prg:/path/to/lb.pl
   RewriteRule   ^/(.+)$ ${lb:$1}           [P,L]
   

   Then we write lb.pl:

   #!/path/to/perl
   ##
   ##  lb.pl -- load balancing script
   ##
   
   $| = 1;
   
   $name   = "www";     # the hostname base
   $first  = 1;         # the first server (not 0 here, because 0 is myself)
   $last   = 5;         # the last server in the round-robin
   $domain = "foo.dom"; # the domainname
   
   $cnt = 0;
   while (<STDIN>) {
       $cnt = (($cnt+1) % ($last+1-$first));
       $server = sprintf("%s%d.%s", $name, $cnt+$first, $domain);
       print "http://$server/$_";
   }
   
   ##EOF##
   

   A last notice: Why is this useful? Seems like www0.foo.com still is overloaded? The answer is yes, it is overloaded, but with plain proxy throughput requests, only! All SSI, CGI, ePerl, etc. processing is completely done on the other machines. This is the essential point.
4. Hardware/TCP Round-Robin

   There is a hardware solution available, too. Cisco has a beast called LocalDirector which does a load balancing at the TCP/IP level. Actually this is some sort of a circuit level gateway in front of a webcluster. If you have enough money and really need a solution with high performance, use this one.

On the net there are a lot of nifty CGI programs. But their usage is usually boring, so a lot of webmaster don't use them. Even Apache's Action handler feature for MIME-types is only appropriate when the CGI programs don't need special URLs (actually PATH_INFO and QUERY_STRINGS) as their input. First, let us configure a new file type with extension .scgi (for secure CGI) which will be processed by the popular cgiwrap program. The problem here is that for instance we use a Homogeneous URL Layout (see above) a file inside the user homedirs has the URL /u/user/foo/bar.scgi. But cgiwrap needs the URL in the form /~user/foo/bar.scgi/. The following rule solves the problem:

RewriteRule ^/[uge]/([^/]+)/\.www/(.+)\.scgi(.*) ...
... /internal/cgi/user/cgiwrap/~$1/$2.scgi$3  [NS,T=application/x-http-cgi]

Or assume we have some more nifty programs: wwwlog (which displays the access.log for a URL subtree and wwwidx (which runs Glimpse on a URL subtree). We have to provide the URL area to these programs so they know on which area they have to act on. But usually this ugly, because they are all the times still requested from that areas, i.e. typically we would run the swwidx program from within /u/user/foo/ via hyperlink to

/internal/cgi/user/swwidx?i=/u/user/foo/

which is ugly. Because we have to hard-code both the location of the area and the location of the CGI inside the hyperlink. When we have to reorganize the area, we spend a lot of time changing the various hyperlinks.

The solution here is to provide a special new URL format which automatically leads to the proper CGI invocation. We configure the following:

RewriteRule   ^/([uge])/([^/]+)(/?.*)/\*  /internal/cgi/user/wwwidx?i=/$1/$2$3/
RewriteRule   ^/([uge])/([^/]+)(/?.*):log /internal/cgi/user/wwwlog?f=/$1/$2$3

Now the hyperlink to search at /u/user/foo/ reads only

HREF="*"

which internally gets automatically transformed to

/internal/cgi/user/wwwidx?i=/u/user/foo/

The same approach leads to an invocation for the access log CGI program when the hyperlink :log gets used.

Here comes a really esoteric feature: Dynamically generated but statically served pages, i.e. pages should be delivered as pure static pages (read from the filesystem and just passed through), but they have to be generated dynamically by the webserver if missing. This way you can have CGI-generated pages which are statically served unless one (or a cronjob) removes the static contents. Then the contents gets refreshed.

RewriteCond %{REQUEST_FILENAME}   !-s
RewriteRule ^page\.html$          page.cgi   [T=application/x-httpd-cgi,L]

Here a request to page.html leads to a internal run of a corresponding page.cgi if page.html is still missing or has filesize null. The trick here is that page.cgi is a usual CGI script which (additionally to its STDOUT) writes its output to the file page.html. Once it was run, the server sends out the data of page.html. When the webmaster wants to force a refresh the contents, he just removes page.html (usually done by a cronjob).

Wouldn't it be nice while creating a complex webpage if the webbrowser would automatically refresh the page every time we write a new version from within our editor? Impossible?

No! We just combine the MIME multipart feature, the webserver NPH feature and the URL manipulation power of mod_rewrite. First, we establish a new URL feature: Adding just :refresh to any URL causes this to be refreshed every time it gets updated on the filesystem.

RewriteRule   ^(/[uge]/[^/]+/?.*):refresh  /internal/cgi/apache/nph-refresh?f=$1

Now when we reference the URL

/u/foo/bar/page.html:refresh

this leads to the internal invocation of the URL

/internal/cgi/apache/nph-refresh?f=/u/foo/bar/page.html

The only missing part is the NPH-CGI script. Although one would usually say "left as an exercise to the reader" ;-) I will provide this, too.

#!/sw/bin/perl
##
##  nph-refresh -- NPH/CGI script for auto refreshing pages
##  Copyright (c) 1997 Ralf S. Engelschall, All Rights Reserved.
##
$| = 1;

#   split the QUERY_STRING variable
@pairs = split(/&/, $ENV{'QUERY_STRING'});
foreach $pair (@pairs) {
    ($name, $value) = split(/=/, $pair);
    $name =~ tr/A-Z/a-z/;
    $name = 'QS_' . $name;
    $value =~ s/%([a-fA-F0-9][a-fA-F0-9])/pack("C", hex($1))/eg;
    eval "\$$name = \"$value\"";
}
$QS_s = 1 if ($QS_s eq '');
$QS_n = 3600 if ($QS_n eq '');
if ($QS_f eq '') {
    print "HTTP/1.0 200 OK\n";
    print "Content-type: text/html\n\n";
    print "<b>ERROR</b>: No file given\n";
    exit(0);
}
if (! -f $QS_f) {
    print "HTTP/1.0 200 OK\n";
    print "Content-type: text/html\n\n";
    print "<b>ERROR</b>: File $QS_f not found\n";
    exit(0);
}

sub print_http_headers_multipart_begin {
    print "HTTP/1.0 200 OK\n";
    $bound = "ThisRandomString12345";
    print "Content-type: multipart/x-mixed-replace;boundary=$bound\n";
    &print_http_headers_multipart_next;
}

sub print_http_headers_multipart_next {
    print "\n--$bound\n";
}

sub print_http_headers_multipart_end {
    print "\n--$bound--\n";
}

sub displayhtml {
    local($buffer) = @_;
    $len = length($buffer);
    print "Content-type: text/html\n";
    print "Content-length: $len\n\n";
    print $buffer;
}

sub readfile {
    local($file) = @_;
    local(*FP, $size, $buffer, $bytes);
    ($x, $x, $x, $x, $x, $x, $x, $size) = stat($file);
    $size = sprintf("%d", $size);
    open(FP, "<$file");
    $bytes = sysread(FP, $buffer, $size);
    close(FP);
    return $buffer;
}

$buffer = &readfile($QS_f);
&print_http_headers_multipart_begin;
&displayhtml($buffer);

sub mystat {
    local($file) = $_[0];
    local($time);

    ($x, $x, $x, $x, $x, $x, $x, $x, $x, $mtime) = stat($file);
    return $mtime;
}

$mtimeL = &mystat($QS_f);
$mtime = $mtime;
for ($n = 0; $n < $QS_n; $n++) {
    while (1) {
        $mtime = &mystat($QS_f);
        if ($mtime ne $mtimeL) {
            $mtimeL = $mtime;
            sleep(2);
            $buffer = &readfile($QS_f);
            &print_http_headers_multipart_next;
            &displayhtml($buffer);
            sleep(5);
            $mtimeL = &mystat($QS_f);
            last;
        }
        sleep($QS_s);
    }
}

&print_http_headers_multipart_end;

exit(0);

##EOF##

The VirtualHost feature of Apache is nice and works great when you just have a few dozens virtual hosts. But when you are an ISP and have hundreds of virtual hosts to provide this feature is not the best choice.

To provide this feature we map the remote webpage or even the complete remote webarea to our namespace by the use of the Proxy Throughput feature (flag [P]):

##
##  vhost.map
##
www.vhost1.dom:80  /path/to/docroot/vhost1
www.vhost2.dom:80  /path/to/docroot/vhost2
     :
www.vhostN.dom:80  /path/to/docroot/vhostN

##
##  httpd.conf
##
    :
#   use the canonical hostname on redirects, etc.
UseCanonicalName on

    :
#   add the virtual host in front of the CLF-format
CustomLog  /path/to/access_log  "%{VHOST}e %h %l %u %t \"%r\" %>s %b"
    :

#   enable the rewriting engine in the main server
RewriteEngine on

#   define two maps: one for fixing the URL and one which defines
#   the available virtual hosts with their corresponding
#   DocumentRoot.
RewriteMap    lowercase    int:tolower
RewriteMap    vhost        txt:/path/to/vhost.map

#   Now do the actual virtual host mapping
#   via a huge and complicated single rule:
#
#   1. make sure we don't map for common locations
RewriteCond   %{REQUEST_URI}  !^/commonurl1/.*
RewriteCond   %{REQUEST_URI}  !^/commonurl2/.*
    :
RewriteCond   %{REQUEST_URI}  !^/commonurlN/.*
#
#   2. make sure we have a Host header, because
#      currently our approach only supports
#      virtual hosting through this header
RewriteCond   %{HTTP_HOST}  !^$
#
#   3. lowercase the hostname
RewriteCond   ${lowercase:%{HTTP_HOST}|NONE}  ^(.+)$
#
#   4. lookup this hostname in vhost.map and
#      remember it only when it is a path
#      (and not "NONE" from above)
RewriteCond   ${vhost:%1}  ^(/.*)$
#
#   5. finally we can map the URL to its docroot location
#      and remember the virtual host for logging puposes
RewriteRule   ^/(.*)$   %1/$1  [E=VHOST:${lowercase:%{HTTP_HOST}}]
    :

How can we forbid a list of externally configured hosts from using our server?

For Apache >= 1.3b6:

RewriteEngine on
RewriteMap    hosts-deny  txt:/path/to/hosts.deny
RewriteCond   ${hosts-deny:%{REMOTE_HOST}|NOT-FOUND} !=NOT-FOUND [OR]
RewriteCond   ${hosts-deny:%{REMOTE_ADDR}|NOT-FOUND} !=NOT-FOUND
RewriteRule   ^/.*  -  [F]

For Apache <= 1.3b6:

RewriteEngine on
RewriteMap    hosts-deny  txt:/path/to/hosts.deny
RewriteRule   ^/(.*)$ ${hosts-deny:%{REMOTE_HOST}|NOT-FOUND}/$1
RewriteRule   !^NOT-FOUND/.* - [F]
RewriteRule   ^NOT-FOUND/(.*)$ ${hosts-deny:%{REMOTE_ADDR}|NOT-FOUND}/$1
RewriteRule   !^NOT-FOUND/.* - [F]
RewriteRule   ^NOT-FOUND/(.*)$ /$1

##
##  hosts.deny
##
##  ATTENTION! This is a map, not a list, even when we treat it as such.
##             mod_rewrite parses it for key/value pairs, so at least a
##             dummy value "-" must be present for each entry.
##

193.102.180.41 -
bsdti1.sdm.de  -
192.76.162.40  -

How can we forbid a certain host or even a user of a special host from using the Apache proxy?

We first have to make sure mod_rewrite is below(!) mod_proxy in the Configuration file when compiling the Apache webserver. This way it gets called before mod_proxy. Then we configure the following for a host-dependent deny...

RewriteCond %{REMOTE_HOST} ^badhost\.mydomain\.com$
RewriteRule !^http://[^/.]\.mydomain.com.*  - [F]

...and this one for a user@host-dependent deny:

RewriteCond %{REMOTE_IDENT}@%{REMOTE_HOST}  ^badguy@badhost\.mydomain\.com$
RewriteRule !^http://[^/.]\.mydomain.com.*  - [F]

Sometimes a very special authentication is needed, for instance a authentication which checks for a set of explicitly configured users. Only these should receive access and without explicit prompting (which would occur when using the Basic Auth via mod_auth).

We use a list of rewrite conditions to exclude all except our friends:

RewriteCond %{REMOTE_IDENT}@%{REMOTE_HOST} !^friend1@client1.quux-corp\.com$
RewriteCond %{REMOTE_IDENT}@%{REMOTE_HOST} !^friend2@client2.quux-corp\.com$
RewriteCond %{REMOTE_IDENT}@%{REMOTE_HOST} !^friend3@client3.quux-corp\.com$
RewriteRule ^/~quux/only-for-friends/      -                                 [F]

How can we program a flexible URL Deflector which acts on the "Referer" HTTP header and can be configured with as many referring pages as we like?

Use the following really tricky ruleset...

RewriteMap  deflector txt:/path/to/deflector.map

RewriteCond %{HTTP_REFERER} !=""
RewriteCond ${deflector:%{HTTP_REFERER}} ^-$
RewriteRule ^.* %{HTTP_REFERER} [R,L]

RewriteCond %{HTTP_REFERER} !=""
RewriteCond ${deflector:%{HTTP_REFERER}|NOT-FOUND} !=NOT-FOUND
RewriteRule ^.* ${deflector:%{HTTP_REFERER}} [R,L]

... in conjunction with a corresponding rewrite map:

##
##  deflector.map
##

http://www.badguys.com/bad/index.html    -
http://www.badguys.com/bad/index2.html   -
http://www.badguys.com/bad/index3.html   http://somewhere.com/

This automatically redirects the request back to the referring page (when "-" is used as the value in the map) or to a specific URL (when an URL is specified in the map as the second argument).