As a quick follow-up to the Opera Turbo testing recently carried out, a piece of open-source software came to my attention that replicates loosely some of the functionality present within Opera Turbo server-side processes.

RabbIT is a web-proxy best implemented at ISP level, since it does not immediately decrease the amount of bandwidth needed to view a given set of web pages. The best way to look at it is that it functions like a regular transparent squid-proxy, albeit with extra provisioning for compressing page elements once downloaded, and then delivering these compressed elements to users.

Because it functions in this manner, RabbIT is really of limited use if implemented at a local server level, unless the local LAN has very little bandwidth. Equally, as most ISPs are unlikely to negotiate and install the software on their equipment on your behalf, a solution must be established whereby its advantages are best utilised.

One model proposed is that a dedicated or VPS server based at a relevant geographical location is purchased. Because hosting companies purchase bandwidth in bulk from upstream telcos, a plan can often be purchased whereby that host is allowed an unlimited amount of data transfer. As an added bonus, since the machine is likely to be housed at a dedicated facility, downstream line speed should be more than generous. RabbIT web proxy would then be installed on this machine with some form of basic authentication.

Back at the client end, individual machines and/or a router is configured with gateway details matching those of  the proxy. When a user browses to a bandwidth heavy site the VPS downloads data, compresses it, and delivers this over the line. I suppose you could term this kind of implementation “cloud caching”. Unfortunately RabbIT is currently limited to only HTTP 1.1 traffic, but there seems to be no reason why it could not work in harmony with a local caching solution such as Squid, which would then serve to keep copies of RabbIT generated pages.

Sure, the extra hop is far from ideal, and the system requirements of RabbIT are fairly beefy, but in some cases this could prove a viable proposition; especially when one considers how costly mobile access methods like satellite and cellular based broadband are, as well as how starved of bandwidth they tend to be. On the plus side, RabbIT is written in Java (a dubious honour) and is multi-threaded unlike many proxy caching solutions.