v6ops and v6 connectivity deployment model

[Pekka Savola <pekkas@netcore.fi>]

Hi,

[ apologies for sending this out with a large distribution but I think
this is issue is about the most significant decision we as the IETF is
going to make about the IPv6 deployment model.  But I think it's important
to bring the issue up with the folks who're thinking about this from a
more generic perspective as well.  ]

Yesterday, Christian Huitema et al. dragged me along to talk about the
issues in unmanaged ("home networks") connectivity in v6ops.

The result was pretty grim .. or bright, depending on where you look at
the problem.  The recommendation is basically:

 - if native available, use it
 - if tunnel service available, use it
 - if 6to4 available, use it
 - if teredo available, use it

However, the current deployment model for Teredo is that every Microsoft 
host implements and enables it.  This would be similar to the situation 
where every v6 host would have to run an "internal" 6to4 relay, or at 
most, have one at the site.

The main point of the automatic mechanisms is that if we buy the
assumption that everyone must be able to turn on v6 (not having to wait
for his ISP), they're basically the only realistic option.

So, this model is basically pushing the responsibility of communicating
with nodes using transition mechanisms to the "properly deployed"
dual-stack systems.

I have big issues with this, but if folks rather want to see v6 happen
instead having it happen "properly", that is one direction to take.  This
is certainly a model where we can say, "v6 is available to [almost
everyone] everyone whenever they want it", regardless of what the ISPs
support.

The main point of this deployment model for something like Teredo is that
it is assumed that the hosts behind NATs *primarily* want to communicate
with other hosts behind NATs (e.g., peer-to-peer apps) -- but other *IPv6*
hosts are not so important.

I'm not sure if I buy these premises either, but if we want to get rid of
about a dozen v4 NAT traversal mechanisms deployed in the Internet, I
guess this kind of deployment model may make sense even if it puts the
burden on folks who have deployed v6 "properly".

Of course, it's also important to consider the end-game, i.e., when the v6
automatic tunneling mechanism "internal relay" support can be removed from
hosts/sites.  Pretty much never in the next 10..15 years, as long as there
are NATs and non-v6 capable ISPs, I guess..

Summary: if you have strong feelings about v6 connectivity deployment
model, I suggest you (and/or someone else) come to the v6ops session
today, 1530-1730.

Note: we didn't originally allocate time for this on the agenda, but we'll 
arrange some -- but we don't have to make a decision about this today..

-- 
Pekka Savola                 "You each name yourselves king, yet the
Netcore Oy                    kingdom bleeds."
Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings