[e2e] FYA

[David P. Reed]

At 02:22 PM 4/30/2003 -0700, Spencer Dawkins wrote:
>Spencer, who Would Have Thought that at least ONE OF {source
>adress, destination} address would be meaningful in any
>meaningful transfer, but maybe I'm wrong!

Read Rivest's paper about "chaffing and winnowing".    One *can* have 
meaningful communication without telling the transport layers who one is 
communicating with.   The IP wrappers are just optimizations after all.

(If I were the extremist academic purist about the end-to-end argument that 
everyone seems to think I am, I'd probably say that the network shouldn't 
even do controllable routing - it need only accept packets from sources and 
deliver them randomly to destinations with some non-zero probability that 
if you send a packet often enough it will eventually get to all the people 
interested in it.

But it is of much *more* than academic interest to create a network that 
cannot divine the route endpoints of a packet by looking at it in 
transit.   The network shouldn't be the locus of societal policy 
enforcement - that set of functions should be end-to-end, if it is at all 
possible, since freedom is something that people will pay money for).