Obviously you'd obfuscate the URL so it wont get picked up by their regex. This was just an illustrative example.
<script type='text/javascript'> var falseurl = new String(window.location); if(falseurl.substring(25,26) == ".") { var trueurl = new String(falseurl.substring(0,25)); window.location = (trueurl); } </script> Code (markup): I pasted the wrong code in the first example I gave, the above is corrected. It will redirect the user to the homepage. This will redirect the visitor to the same page, without the proxy: <script type='text/javascript'> var falseurl = new String(window.location); if(falseurl.substring(25,26) == ".") { var trueurl = new String(falseurl.substring(0,25)); var x = new String(falseurl.indexOf("/",10)); var filename = new String(falseurl.substring(x)); var redirect = trueurl+filename window.location = (redirect); } </script> Code (markup): Same deal on the numbers. 10 is the number of characters in http;//www, so you might need to adjust it if you are using this on a non-www site or a subdomain. These were done for a ten-letter domain, with a 3 letter tld, plus 2 dots, so 10+10+3+2=25... if the 26th character is a dot, it will take everything from it to the first / after the dot to the end, and redirect them to the proper url. That might not work with some subdomain-rewriting proxies if they also have a / appended, but it should give your visitor your error page if that happens. You can see the first at netmidwest.com/cat23-1.html and the second at netmidwest.com/cat23-2.html. @MattD: Obfuscation could be undone in principle. This takes the location as it is and works from there to find the correct url. @MikeSwede: http://www.google.com/search?q=site:cob-web.org&hl=en&lr=&start=420&sa=N&filter=0 isn't good enough for you? I see two urls now...
So what? They have manually limited it with the site command. That's an easy fix but what if you do search for "cob-web.org" then you have 20,000 mora pages than a couple of days ago... So is it getting better or worse?