Re: [RRG] Migration to IPv6-only addresses

[Robin Whittle <rw@firstpr.com.au>]

Hi Brian,

You wrote:

RW> I can't avoid the conclusion that we are likely to be stuck with
RW> IPv4 and NAT forever.  I would be happy for someone to prove me
wrong!
> 
> The model has always been to plan for an indefinite period of
> co-existence, with the background assumption that after some
> large number of years, IPv4-only devices would die a natural
> death. Whether you include that assumption or not has no
> practical impact today in terms of what we should do. 

I should have written "several decades" rather than "forever".

The main task of the RRG is to choose some architectural proposals -
probably for improving BGP and for a new IP-level ITR-ETR tunneling
system for both IPv4 and IPv6 - to be developed and widely deployed
in a timeframe of a few years.

> I'm not sure that discussion beyond that point belongs on RRG.

I believe it is pertinent, since I sense a number of folks here -
perhaps including yourself - seem to think that the solution is to
migrate to IPv6.

In the foreseeable future - say ten years or more - I can't see how
anyone would pay for an Internet service which had only an IPv6
address.  (A likely exception is a cell-phone, which has highly
restricted communications abilities and in which the application
software is usually not chosen or provided by the user.)  It would
not be useful in general, since only a handful of protocols could be
supported via proxies to communicate with the rest of the Net which
uses IPv4.


I really appreciate you pointing out the errors of some proposals,
such as the use of option headers.

I understand that not everyone has the time or the inclination to do
so, but what I would really appreciate is a critique of Ivip.  The
I-D has been around for over two months now.  I have provided what I
intend to be constructive critiques of the other proposals, apart
from TRRP.

I don't think it is reasonable to talk up IPv6 as a solution to the
problems of the next 5 to 10 years:

  1 - IPv4 address depletion (possible solution: more efficient
      use of space, for more actively used IP addresses and/or
      more user networks, perhaps with smaller numbers of
      addresses than existing BGP arrangements can handle).

  2 - Minimising growth in global routing table vs. supporting
      more and more end-user networks for multihoming and for
      choice of ISP without renumbering.

  3 - Robustness and convergence time of the BGP system.

IPv6 is just as bad as IPv4 for 2 and 3.  Since IPv4 was not
designed with an upgrade strategy, IPv6 doesn't help with 1 either -
because having an IPv6-only address will provide a far smaller
subset of connectivity (in terms of protocols, applications and/or
reachable hosts) than almost anyone would find acceptable.

There are a bunch of people trying to develop IP-based ITR-ETR etc.
solutions which will actually help with the immediate problems 2 and
3 above.  I intend Ivip to help with 1 too.  I think they would all
help with 1 - by enabling more effective use of IPv4 address space.
 Dino seems to agree that LISP would help to some extent with this:

  http://psg.com/lists/rrg/2007/msg00368.html

What I think is really needed is constructive critiques of the
current ITR-ETR proposals.  If you (and others) think they won't
work at all, then please say so.  If you (and others) think they
have some chance of working, then please say so, and point out their
faults or how they might be improved.

I think it was you who pointed out a fundamental flaw in LISP-NERD -
that all its ITRs are caching ITRs, reliant on DNS lookups, which
take a second or more (perhaps several seconds) before they are able
to tunnel a packet to the correct ETR.  This means that the whole
thing will not work since TCP SYN and/or SYN-ACK packets (depending
on whether the sending and/or the receiving host uses
LISP-NERD-mapped addresses) will be delayed by so long that the host
software will give up, or (I guess) send multiple retries, or try
some other host.  Similar delays would probably clobber or disrupt
UDP-based protocols.  If they didn't stop the protocols from
working, they would certainly slow things down unacceptably.

I agree entirely with this critique.  I think this reliance on DNS
and caching makes LISP-NERD and TRRP unworkable.

That leaves LISP-CONS, eFIT-APT and Ivip.

What I am suggesting is that if you or anyone else has time, it
would be helpful for you to discuss whatever you think is good or
bad about the current ITR-ETR proposals regarding IPv4.

In the next five or ten years, I believe IPv6 is at best going to be
an adjunct to IPv4, rather than a replacement for it.  I think this
is what you wrote in your last message.  If anyone can show why
almost all hosts (3G cellphones and experimental, special-purpose
hosts excepted) will not need to communicate directly with IPv4
hosts in the next 10 years, then please do so.

I don't think it is good enough to portray what a screaming disaster
IPv4 will be as an argument for why enough people will soon jump
ship and adopt IPv6.  Humans have an extraordinary capacity to
battle on through incrementally worsening disasters - including
especially in IT when everyone is dependent on an existing system.

For IPv6 to be part of the solution to the current crisis, there
would have to be a way that ordinary users will be happy to have
IPv6-only addresses, in the 2010 to 2015 timeframe.  (Unless it
could be shown that a subset of users with IPv6 addresses will
appreciably reduce the load on the IPv4 system.) Its no fault of
IPv6 that this is a difficult or impossible task.

  - Robin


--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg