I see many different tools out there, but none seem to be canonical. I would also prefer open source, if possible (or even start an open source project), as it may be possible to contribute back improvements.

Some notable examples I will be trying to emulate are: http://www.google.com http://www.msn.com

If you have any experience with this problem, please speak out! I'd love to hear about your experiences. I especially would appreciate information on how to ensure that the resulting output file is identical (functionally) to the input file and how to gauge how effective the compression was. (One has to take into account client-side caching and JS-processing time over a pure size comparison.)

Thanks!