You're right that a flat mesh is not the best topology for
long-distance communication, especially with current routing
protocols, which require things like global lists of all routeable
prefixes.
On the protocol front, I suggest that the IETF develop routing
protocols that can work well in a flat mesh topology. Parallelizing
traffic streams over many available routes, so it all doesn't try to
take the shortest path, would appear to be a particularly important
feature, as would preventing all of a node's links from being swamped
by through traffic that other nodes want it to route.
The problem with long-distance traffic over flat mesh networks is less
with throughput (if everything isn't taking the shortest path) than
with the latencies involved in sending traffic over a very large
number of hops. I think the solution there is to send traffic that's
leaving your local area over the existing (tapable) long-distance
infrastructure. The idea is to make tapping expensive, not impossible.
There's also the point to be made that current traffic patterns depend
to a significant extent on current Internet architectural decisions.
If everyone had a gigabit connection to their neighbors, but only a 10
megabit uplink to route long-distance traffic over, they might find a
use for all that extra local bandwidth.
On Fri, Sep 6, 2013 at 7:22 AM, Noel Chiappa
<jnc(_at_)mercury(_dot_)lcs(_dot_)mit(_dot_)edu> wrote:
> One way to frustrate this sort of dragnet surveillance would be to
> reduce centralization in the Internet's architecture.
> ...
> [If] The IETF focused on developing protocols (and reserving the
> necessary network numbers) to facilitate direct network peering between
> private individuals, it could make it much more expensive to mount
> large-scale traffic interception attacks.
I'm not sure this is viable (although it's an interesting concept).
With our current routing tools, switching to a flat mesh, as opposed to the
current fairly-structured system, would require enormous amounts of
configuration/etc work on the part of smaller entities.
Also, traffic patterns being what they are (e.g. most of my traffic goes
quite a distance, and hardly any to things close by), everyone would wind up
handling a lot of 'through' traffic - orders of magnitude more than their
current traffic load.
Noel