[e2e] Overly Overlay; Peer to peer is commonplace

J. Noel Chiappa jnc at ginger.lcs.mit.edu
Mon Dec 31 14:48:58 PST 2001

    > From: "David P. Reed" <dpreed at reed.com>

    > meanwhile, the end-to-end protocol developers seem to be building into
    > their models of the network various pernicious assumptions

An extremely good idea to note this, and a generally good list of points; with
one thing that flagged my attention:
    > - that IP addresses should reflect topology because topology is stable,
    > because putting in fiber requires backhoes,

I am reminded of a wonderful quote by Earl St Vincent, who said of Napoleon
and the French: "I do not say they cannot come; I only say they cannot come
by sea."

If you want to have a namespace for communicating entities, where the names do
not depend on the location of the entities in the network, I'm all for that.
(In fact, see: http://users.exis.net/~jnc/tech/endpoints.txt for advocacy of
such a thing.)

However, to say that "routing-names" (and, like it or not, IPvN addresses are
currently routing-names, in addition to other functions) in a very large
network should *not* reflect network location (i.e. topology) is seriously
confused. (Technical justification left out because I don't want to append a

Yes, ad-hoc radio networks are a special case, and require some special
measures to handle them, particularly in light of their highly dynamic
topologies. What the localized routing looks like inside such a system, and
how that would impact on the names that localized routing might use, are
interesting questions.

(I think RFC-2103 has something to say about the interaction of routing-names,
and highly dynamic mobile networks, such as packet radio, BTW.)

However, that does not change the requirement that the routing-names of the
entities in such a local network - at least, the routing-names with
*system-wide* scope/visibility/utility - need to say something about where in
the overall Internet they are.

In other words, "I do not say you cannot have names which do not reflect
topology; only that the routing calculations cannot use those names."


More information about the end2end-interest mailing list