[e2e] Revisiting RON ("traffic engineering considered harmful"

[Joe Touch]

David P. Reed wrote:

> IMHO, we wouldn't need "overlay networks" if we could get over the idea 
> that IP by definition implies BGP or some other "universal solvent for 
> routing".


It depends on what the overlay is for; I believe overlays are still 
useful, even if the underlying routing is OK, for experiments, testbeds, 
teaching labs, and especially to connect sets applications.

The latter case is a network-layer version of peer networks, which 
avoids recapitulating the network layer inside the application.

FWIW, we have an overlay project as well, the X-Bone, with further 
information at http://www.isi.edu/xbone , available in 
/usr/ports/net/xbone on FreeBSD and as a Linux RPM from our website.
The X-Bone focuses on how to do an overlay using existing IP, IPsec, 
etc. protocols (we're working on a draft for the upcoming IETF on this).

(PS - we're especially interested in helping it be used for teaching 
labs as well)


> It would be much more interesting to focus less on "global optimality" 
> (of the sort that BGP strives for) and more on heterogeneous routing 
> approaches co-existing.


See X-Bone :-) We support using existing routing protocols on the overlay.

Joe