Jeremy Bowers makes a number of good points about how current usage of RSS is pretty wasteful of bandwidth. He even proposes using rproxy to improve the situation.
Sometimes I think this is an interesting problem to tackle. Yesterday, I was cooking up fancy P2P schemes for scalable, persistent distribution of RSS fragments. Not an original idea, but just something to kick around.
Then I remember that a) RSS traffic is still ridiculously small relative to regular HTTP traffic and b) part of RSS's popularity is that it does just use HTTP.
It's a format optimized for publishers not consumers, so if you want to fix any problems you'll really have to start at that end of the pipe.
Thanks to Ted Leung