Re: Enterprise Analysis DSTM Issue

[Fred Baker <fred@cisco.com>]

On Aug 9, 2005, at 2:13 PM, Tim Chown wrote:
> <devil's advocate>
> What does the IETF recommend for dual-stack nodes in an IPv6  
> infrastructure?
> </devil's advocate>

The IETF's current recommendations are encapsulated in http:// 
www.ietf.org/rfc/rfc2893.txt. The abstract is:

   This document specifies IPv4 compatibility mechanisms that can be
   implemented by IPv6 hosts and routers.  These mechanisms include
   providing complete implementations of both versions of the Internet
   Protocol (IPv4 and IPv6), and tunneling IPv6 packets over IPv4
   routing infrastructures.  They are designed to allow IPv6 nodes to
   maintain complete compatibility with IPv4, which should greatly
   simplify the deployment of IPv6 in the Internet, and facilitate the
   eventual transition of the entire Internet to IPv6.  This document
   obsoletes RFC 1933.

As Jim is wont to point out, this doesn't address the case of an IPv6- 
only infrastructure. The deployment of IPv6-only infrastructures  
greatly complicates the transition, as we have no guarantee that  
there is a way to bridge IPv6 deployed over any given IPv4  
infrastructure, and the presence of IPv6-only infrastructures means  
that we can no longer make that guarantee for IPv4/IPv6 either. In  
addition, given that there are a variety of different transition  
models, many of which similarly pick off a piece of the puzzle  
without analyzing its effectiveness end to end:

        draft-blanchet-v6ops-tunnelbroker-tsp-02.txt
        draft-despres-v6ops-transition-v5roadmap-00.txt
        draft-huitema-v6ops-teredo-05.txt
        draft-kim-v6ops-ipv6overwibro-issues-00.txt
        draft-liumin-v6ops-silkroad-03.txt
        draft-massar-v6ops-heartbeat-01.txt
        draft-massar-v6ops-tunneldiscovery-00.txt
        draft-ooms-v6ops-bgp-tunnel-05.txt

Even if every IPv4-only or IPv6-only network provides a way to bridge  
the other over it, I'm not aware of a proof that end-to-end  
connectivity is eventually assured, as I don't have a guarantee that  
they will not touch each other and (as would be required) provide all  
relevant transition mechanisms over all relevant paths. There is a  
good article on this concept at http://www.isen.com/papers/ 
Dawnstupid.html. Yes, you heard this from someone whose day job is  
with a company that likes to talk about the "Intelligent Information  
Network". Let's just say that I don't work in marketing...

Oh, by the way, as IPv4 address space allocation in fact approaches  
impossibility, the deployment of IPv6-only infrastructure is not only  
predictable, it is unavoidable. At this point, it is very avoidable,  
and I have a hard time arguing that the complications resulting from  
it are preferable to the complications of running a dual stack network.

I'm all for lrw, v6tc, or whatever it winds up being called. I was  
last time around as well. AFAIK, v6tc is *also* not chartered to  
solve the entire complexity of the end to end problem - it is  
chartered to come up with a rendezvous protocol for a tunneling  
solution, and presumably not a dozen mutually incompatible ones. From  
what Mark Townsley and Dave Ward told me, something like DSTM might  
be part of that, but it has to manage an arbitrary underlying tunnel  
architecture - MPLS, ATM, L2TP, IP/IP6, GRE6, etc, and perhaps a  
couple of such strung together end to end - and is therefore probably  
not exactly DSTM. But DSTM would be a useful input. I don't think we  
should be deploying IPv6-only networks in any form until we have  
figured out the end to end problem.