Since filtering the pages is a "heavy" process, RabbIT caches the pages it filters while trying to respect things like pragma no-cache. RabbIT also accepts request for nonfiltered pages by prepending "noproxy" to the adress (like http://noproxy.www.altavista.com/). Optionally, a link to the unfiltered page can be inserted at the top of each page automatically.
RabbIT is developed and tested under Solaris and Linux. Since the whole package is written in java, the basic proxy should run on any plattform that supports java. Image processing requires a program called convert (found in ImageMagik). RabbIT can be run without image processing enabled, but that loses part of the idea.
RabbIT works best if it is run on a computer with a fast link (typically your ISP). Since every large image is compressed before it is sent from the ISP to you, surfing becomes much faster at the price of some decrease in image quality. If some parts of the page are already cached by the proxy, the speedup will often be quite amazing. For 1275 random images only 22% (2974108 bytes out of a total of 13402112) were sent to the client.
RabbIT works by modifying the pages you visit so that your browser never sees the advertising images, it only sees one fixed image tag (that image is cached in the browser the first time it is downloaded, so sequential requests for it is made from the browsers cache, giving a nice speedup). For images RabbIT fetches the image and run it through a processor giving a 10% jpeg instead of the animated gif-image. This image is very much smaller and download of it should be quick even over a slow link (modem).