#!/usr/bin/perl
#
# log-tail, (c) Andrew Daviel <andrew (at) vancouver-webpages(dot)com>
# Vancouver Webpages, 1996
# http://vancouver-webpages.com/proxy/log-tail.pl
# You may do what you like with this except say that you wrote it
# or sell it for profit.
#
# This program is invoked as a CGI script
# URL such as http://somewhere.org/cgi-bin/log-tail/error_log
# Although it does print out the last entries in logfiles, its
# primary purpose is to demonstrate a method of generating
# a cacheable CGI script. 

# Modern browsers use local cache extensively, and a multi-level
# national and international cache hierarchy is being developed
# (see http://www.nlanr.net/Cache/, http://vancouver-webpages.com/proxy/
# for more information).
# Where objects can be cached the response time is greatly improved,
# which is why Netscape is better than the original Mosaic over
# an analog modem. Hierarchical cache allows popular documents to be
# cached nearby, so that they may be accessed at LAN speeds instead of
# the (sometimes pitiful) national/International Internet speeds
# 
# How documents are cached depends on the HTTP headers (see the 
# HTTP specification at ftp://ds.internic.net or http://www.w3.org)
# It would be possible to create a document whose content depended
# solely on its URL, which could be cached indefinitely and would
# never expire (something like cgi-bin/add?2+3 giving "5"), but most
# Web pages are generate from a file; the modification time of the file
# is used to generate a Last-Modified header. The Get If-Modified-Since
# mechanism used by Netscape allows the browser to quickly check that
# its locally cached pages and images are up-to-date; where these
# are generated by a CGI script, usually the Last-Modified header is missing
# and a full GET is required, slowing the response time and consuming
# long-distance bandwidth.
# This script demonstrates a technique for generating HTTP headers to
# allow CGI documents to be sensibly cached.

# load the HTTP date routines
$libloc = "/usr/local/lib/libwww-perl-0.40" ; unshift(@INC, $libloc); 
require "wwwdates.pl" ;

# Given a URL such as "http://somewhere/cgi-bin/log-tail?error_log",
# this code redirects it to "http://somewhere/cgi-bin/log-tail/error_log".
# (or perhaps "http://somewhere/cache-cgi/log-tail/error_log".
# This may be used to create a response URL cacheable by the 
# default configuration of the popular Squid proxy-cache agent.
# The original URL with a query term (?)
# might be generated by an HTML form, while the redirected URL can
# be bookmarked and cached, with proper handling of If-Modified-Since.
# This kind of technique can be used to make a hit counter for an image  - 
# the image remains cacheable and the redirect (a few hundred bytes) is counted.

$path = $ENV{'PATH_INFO'} ; $query = $ENV{'QUERY_STRING'};
if ($query && !$path) {
  # figure out our IP address in a semi-portable way. Hard code it if required.
  $host = `hostname` ; chop($host) ;
  print "Content-type: text/html\nLocation: http://$host/cgi-bin/log-tail/$query\n\n";
  exit ;
}

# check that there really is an argument to the script
# An illegal expiry date of "0" ensures that this response is not cached

unless ($path) {
  print "Content-type: text/plain\nExpires: 0\n\n";
  print "You must give a filename in the path, e.g. cgi-bin/log-tail/error_log\n";
  exit ;
}

# Get the value (if any) of the If-Modified-Since header. Netscape operates
# as follows (depending on the cache options set):

# Operation     Cached locally      HTTP transaction
# follow link      No                GET
# follow link      Yes               none (use locally cached copy)
# Reload           No                GET
# Reload           Yes               GET + If-Modified-Since + pragma: no-cache
# Shift-Reload     Yes               GET + pragma: no-cache

# Netscape sends an IMS value such as "Friday, 20-Sep-96 07:26:24 GMT; length=1539"
# i.e. it appends the locally cached length. This script ignores the 
# length, but must strip it for the date routines. The IMS value is converted
# to an HTTP time value (RFC1123)

$imsd = $ENV{'HTTP_IF_MODIFIED_SINCE'} ; 
$ims = $imsd ; $ims =~ s/;.*// ; # for Netscape
$ims = &wwwdates'get_gmtime($ims) ;

$day = 24*60*60 ; $week = $day*7 ; $hr8 = 8*60*60 ;

# add the argument (file name) to the logfile root
$file = "/usr/local/etc/httpd/logs".$path ;

# get the modification time of the file
# For a database lookup, it may be possible to assign a modification
# time to an individual key. Otherwise the modification time of the 
# database may be given.
# For a data acquisition application, such as a Webcam or rain gauge, it
# may be possible to assign an arbitrary modification time based on the rate
# of change of the quantity being measured, perhaps 10 minutes in the past
# for the Webcam. The raingauge might register a change for every millimeter
# of rainfall. Remember that a new current reading may be obtained by
# doing a Reload (i.e. GET without IMS).

@stat = stat($file) ; $size = $stat[7]; $modtime = $stat[9] ;
# also take note of the modification time of the script itself
@stat = stat($0) ; $mymodtime = $stat[9] ;
if ($mymodtime > $modtime) { $modtime = $mymodtime ; }

# If the file has not been modified since the IMS, return a code 304
# "not modified". This syntax works with the later Apache servers.

if ($ims && $modtime<=$ims) {
  print "Status: 304 Not Modified\n\n" ;
  exit ; 
}

# For demonstration purposes, assign an expiry date 5 minutes in the future.
# For a real application, perhaps you know when the database will be
# updated and can generate an appropriate value. In the absence of an Expires
# date, cache agents guess an expiry date based on how recently the document
# was modified, the type of document, etc. - a GIF that hasn't changed in
# 3 months might be cached for a month, though as soon as someone does
# a "reload" it will be updated in cache.

$expires = time + 5*60 ; # expire in 5 minutes

# Convert expiry and last-modified times to RFC1123 time

$gexpires = &wwwdates'wtime($expires,'GMT');
$glastmod = &wwwdates'wtime($modtime ,'GMT');
$expires = &wwwdates'wtime($expires,'');
$lastmod = &wwwdates'wtime($modtime ,'');
$date = &wwwdates'wtime(time,'') ;
$gdate = &wwwdates'wtime(time,'GMT') ;

# This script generates a valid Content-length header.
# This could be used to check against the Netscape IMS value,
# but is otherwise not much used. Certain cache agents may
# refuse to cache documents larger than a certain size.
# Since the content-length is also included in the document
# body, a dummy value is used, then overwritten by the real value.
# A link to the Netscape "about:document" URL is provided
# for convenience. In a real application the HTTP headers would
# not be given in the body (and only GMT values need be calculated)

$head = "<html><title>Last 10 entries in logs$path</title></head>
<body><h1>Last 10 entries in logs$path</h1>
A demonstration of a cacheable CGI script.<br><address>
<a href=\"mailto:andrew@vancouver-webpages.com\">Andrew Daviel</a>,
<a href=\"http://vancouver-webpages.com\">Vancouver Webpages</a></address>
<p><a href=\"about:document\">about:document</a> - about this document (in Netscape)<br>
<p>
Content-type: text/html<br>
Content-Length: xxxxxxx<br>
Expires: $gexpires<br>
Expires: $expires<br>
Last-Modified: $glastmod<br>
Last-Modified: $lastmod<br>
Date: $gdate<br>
Date: $date<br>
If-Modified-Since: $imsd<br>
<p>
Last 10 entries in $file<br>
<pre>
" ;
# generate the tail of the file given as an argument, add it to the
# header and calculate the length.

$tail = `tail -10 $file` ;
$doc = $head.$tail."\n" ;
$length = length($doc) ;
$slength = substr($length."       ",0,7) ;
$head =~ s/xxxxxxx/$slength/ ;
$doc = $head.$tail."\n" ;

# Generate the complete HTTP response for a successful GET
print "Content-type: text/html
Content-Length: $length
Expires: $gexpires
Last-Modified: $glastmod

$doc";