[e2e] Re: End-system routing control (Was: Open the floodgate)

[Jon Crowcroft]

In missive <20040423171956.GD35199 at lcs.mit.edu>, "David G. Andersen" typed:

 >>>   *> discussed on this list before (i think it was JMS who put it that the
 >>>   *> real purist end2end thing would be to take routing decisions out of the
 >>>   *> net and put them in the end systems too along with error recovery) -
 
 >>> that is a correct deduction.  Datagram routing cannot be done
 >>> effectively by end systems alone (you encounter problems of robustness
 >>> and scalability), so it is not really a candidate for movement up the
 >>> stack and towards the edges.

of course the big problem is that we have a network which runs a proprietary protocol 

IPv4 is "cisco proprietary" in the sense that 
the community cannot change IPv4 due to the investement 
we've made by mostly buying routers from a single source
investing them with inventives to optimize arount that stable 
market - (yes, i realize the network effects also mean that 
even if cisco wanted to change it, it would be against shareholder
arguments)

as mark handley said about viruses and operating systems, 
a small gene pool is a recipe for vulnerability - same
lack of flexibility means we cannot innovate in the net 
to evolve to meet new challenges there too...

one solution is to recognize the phases that systems like this go through and regulate (or even 
nationalise)

another is to bypass the net and go innovate around the edges...i am afraid that overlays don't work to do that
though (nor was that the intention of the suggestion that
people work on overlays in the "looking over the fence" report btw, meant to replace real innovation in network
layers - at least 1 person from the committee told me that anyhow - overlays were meant to stimulate higher level
work, and early prototypes of lower layer work which should then move to its rightful place in the
link/net/transport when ready  - we have a massive barrier to doing that today)



j.