[ppml] 2005-1 alternatives

Owen DeLong owen at delong.com
Wed Apr 20 03:26:11 EDT 2005


> But then the duality and implicit tensions of routing scaleability and
> addresses utility goes back a very long way - the Routing and Addressing
> Group of the IETF in the early 1990s was an early incarnation of the same
> set of tensions relating to what makes routing scale vs what makes
> addresses truly useful and convenient to use.
>

That is why I am becoming progressively more convinced that we need
to separate these two functions.  Addressing primarily provides an
end-system identifier function.  Unfortunately, because we have chosen
to also use a portion of this Address to provide routing information as
well, we have painted ourselves into a corner on these issues, repeatedly.

I think it is time to take a step back and consider a more radical approach
to separating these two functions.  In the long run, I think we would do
well to consider the possibility of eliminating prefixes from interdomain
routing, and, instead, build a facility for looking up the origin-AS for
a given prefix, and, routing strictly on origin-AS.  WIthin a given AS,
routing would occur as it does now, but, Interdomain routing is where
we face table overflow, not intr-as routing.

The bottom line is that IPv6 doesn't provide any benefits over IPv4 to
most adopters today.  IPv4 has provisions for getting PI space, and,
people are using them because there are lots of things that just don't
work right or sufficiently without it.  IPv6 has not solved ANY of these
problems.

There are, as I see it, two solutions on the horizon for this situation in
IPv6.  The first is ULA, which, has most of the drawbacks of RFC-1918
address space, but, lacks the self-enforcing limits of RFC-1918 IPv4
address space.  The reality is that absent a forcing function to keep
ULA from getting routed globally (such as the inherent overlap issues
with RFC-1918 space), market demands will become a forcing function
driving more and more ISPs to route more and more ULA space.  Absent
this policy, or, one like it, there are three possible outcomes:

	1.	IPv6 adoption will continue to be delayed until such
		time as IPv4 space is virtually exhausted and there is
		no further possibility to limp along with IPv4.

	2.	IPv6 adopters will drive the global routing of ULA
		over time.

	3.	Most IPv6 adopters will find or fabricate a way to
		act as an LIR out of desperation to obtain PI space
		under existing policy.

	4.	All of the people who have very legitimate and well thought
		out needs for PI address space will simply fade into the
		woodwork and accept IPv6 as is with PA space and renumber
		whenever they have to.

Of the 4, I can see any or some combination of all of the first 3 as being
likely.  I think that the fourth is what the current IETF proposals are
banking on, and, I think that any real-world perspective will clearly
show that there is absolutely no historical precedent or reason to believe
that it is at all likely.

Owen


-- 
If this message was not signed with gpg key 0FE2AA3D, it's probably
a forgery.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 186 bytes
Desc: not available
URL: <https://lists.arin.net/pipermail/arin-ppml/attachments/20050420/553e9e6f/attachment.sig>


More information about the ARIN-PPML mailing list