A Robust Link-Translating Proxy Server Mirroring the Whole Web
- Ziqing Mao ,
- Cormac Herley
MSR-TR-2008-165 |
Link-translating proxies are widely used for anonymous browsing, policy circumvention and WebVPN functions. These are implemented by encoding the destination URL in the path of the URL submitted to the proxy and rewriting all links before returning the document to the client. Commonly these are based on the CGIProxy implementation or a variant. While popular, broken links and very slow load times are common. This is so, since the use of scripting languages makes finding and translating links (the essential task of such a proxy) very difficult. Some web-sites become entirely non-functional when loaded. This paper proposes a novel architecture for a link-translating proxy. By using a sub-domain mapping technique we entirely eliminate the need to translate (or even find) relative links in content. This makes the link rewriting algorithm very simple and robust. We have implemented, deployed, and tested extensively. We conducted a comparison between the our proxy and CGIProxy. Both the theoretical analysis and experimental evaluation show that the our proxy is significantly better in terms of robustness, performance, and security. We call our system ABOProxy, since it involves the address bar only. Our implementation is running and can be used to navigate at will by appending url .aboproxy.com to the hostname of any URL to be loaded.