boston-glob.rb

#!/usr/bin/env ruby -w

require 'cgi'
require 'net/http'
require 'uri'

# Method 1:
# Rewrite article links to Google 'I'm Feeling Lucky' search.
# Site will show article content if referer(sic) is Google.
# Downside: breaks navigation (other links on site lead back to original bostonglobe.com)
REWRITE_METHOD = 'google'

# Method 2:
# Rewrite all links back to this proxy.
# Site will show article content if user-agent is Googlebot.
# Downside: breaks some JavaScript, maybe some layout.
#REWRITE_METHOD = 'proxy'

cgi = CGI.new

myUrl = 'http://' + ENV['SERVER_NAME'] + ENV['REQUEST_URI'].split('?')[0]
baseUrl = 'http://www.bostonglobe.com'
imFeelingLuckyUrl = 'http://www.google.com/search?q=%s+site:bostonglobe.com&btnI=3564'
googlebotUserAgentString = 'Googlebot/2.1 (+http://www.googlebot.com/bot.html)'

if (defined? cgi['path']) && (!cgi['path'].empty?) then
  fetchUrl = baseUrl + '/' + cgi['path']
else
  fetchUrl = baseUrl + '/'
end

url = URI.parse(fetchUrl)
html = String.new
res = Net::HTTP.start(url.host, url.port) do |http|
  response = http.get(url.path, { 'User-Agent' => googlebotUserAgentString })
  
  if (response.code == '200') then
    html = response.body
  else
    cgi.out { "Response %s received from server" % response.code }
    exit
  end
end

# HACKITY HACK HACK!

# Rewrite CSS to absolute URLs
html.gsub!(/(<link[^>]+href=['"]+)((?!http)[^'"]+)/i, '\1' + baseUrl + '\2')

# Rewrite images to absolute URLs
html.gsub!(/(<img[^>]+src=['"]+)((?!http)[^'"]+)/i, '\1' + baseUrl + '\2')

# Rewrite JavaScript to absolute URLs
html.gsub!(/(<script[^>]+src=['"]+)((?!http)[^'"]+)/i, '\1' + baseUrl + '\2')

# Add back fonts CSS... JavaScript is supposed to do it, but it fails due to relative URLs that we didn't catch
fontsCssUrl = 'http://www.bostonglobe.com/css/globe-fonts.css,globe-comments.css'
fontsCssTag = '<link rel="stylesheet" type="text/css" href="%s" />' % fontsCssUrl
html.gsub!(/<\/head>/i, fontsCssTag + '\0')

if (REWRITE_METHOD == 'google') then
  html.gsub!(/(<a[^>]+href=["'])([^'"]+\/)([^\/'"]+)(\/[^\/'"]+\/story\.html)/, '\1' + (imFeelingLuckyUrl % '\3'))
elsif (REWRITE_METHOD == 'proxy') then
  html.gsub!(/(<a[^>]+href=["'])(\/[^'"]*)/) do |s|
    $1 + myUrl + '?path=' + CGI::escape($2[1..-1])
  end
else
  raise "unknown REWRITE_METHOD"
end

cgi.out('charset' => 'utf-8') { html }