Hi all, I've just been catching up, and I've noticed the reappearance of
a common theme that I think warrants discussion directly -- what
difference we should make, how to do it, and if it's even possible. Here
are my thoughts on the issue, many pieces of which have been said by
others before me (Tony, Lixia, Christian, Noel, Dave Conrad, and Brian,
to name a few).

There are many tiny changes happening constantly in the Internet on a
small scale. These changes have the potential to add up to a radically
different picture sometime down the road, as we've seen happen with NAT.
So, to have any influence on the network, we should be thinking about
how we can tweak the small changes to push things in a good direction.

I believe this means a focus on three principles:

(1) The network will always be heterogeneous

(2) Any large change will only happen as the aggregate result of many,
many tiny changes at individual networks and hosts

(3) It's not up to us (or any central authority) which tiny changes get made


Thus, we can't rely on anyone deploying any particular scheme, but we
can provide the tools that will allow for a future where routing is
scalable and robust.

First of all, we need to push forward with ISP-only solutions. It's the
ISPs that stand to suffer the most from a rapidly growing global routing
table, and they need some tools available to them *that do not depend on
end host involvement* in case it gets out of hand.

Second of all, (1) and (2) imply that we should push forward with some
host-based scheme that separates the host's identifier from its routable
address. This is orthogonal to ISP-based solutions, and it is bound to
be just what *someone* needs, and could make some difference in the size
of the routing table. But since it doesn't address the problem directly,
it's not enough on its own.

(2) implies that we need a step-by-step plan to deploy any scalability
solution, which allows for baby steps all the way from the current state
of the network to some reasonable endpoint. (3) implies that this all
has to be possible without anyone mandating anything. We have been
working towards this goal with APT-BGP, and Jari suggested that all
proposals come up with a similar story.

So I think our job lies in deciding which is the best tool for each job,
and the development and promotion of those tools. This means not
focusing on how to create the glorious new, scalable Internet routing
system, but on what tools are needed to prevent skyrocketing ISP costs,
prevent the fragmentation of the network, and any other foreseeable
disasters.

-Michael

_______________________________________________
rrg mailing list
[email protected]
https://www.irtf.org/mailman/listinfo/rrg

Reply via email to