Re: Votes against what? [ draft-arkko-ipv6-transition-guidelines WGLC]

[Cameron Byrne <cb.list6@gmail.com>]

On Fri, Aug 20, 2010 at 10:01 PM, Fred Baker <fred@cisco.com> wrote:
>
> On Aug 20, 2010, at 9:11 PM, Randy Bush wrote:
>
>>> Actually they are votes
>>
>> there are?  votes?  i thought we had discussions on the merits of
>> various engineering approaches.
>
> These aren't IETF votes, Randy. These are network operators voting with their hard-earned cash on how to deploy.
>
> There are of course those that are not deploying. There is not much to be said about them other than that.
>
> People that are doing green field networks may be implementing IPv6-only. I have heard several mobile operators suggest it, and I know one that has. I know of one research network, CERNET2, that has also done so.
>

FYI, FWIW, my IPv6-only service is deployed on a network that is not
greenfield.  More on that below.

> Those that have existing IPv4 networks, to my knowledge, are not first shutting down their IPv4 networks and then bringing up IPv6; they are bringing ip IPv6 in the IPv4 networks.
>

FYI, FWIW, i am also not shutting down IPv4.  My approach is IPv6-only
deployed one handset model at a time.... no disturbance to existing
install base.... no boiling of the ocean.

> You would be a well-placed party to observe on this. What incidence have you seen of IPv6-only deployment, and what incidence have you seen of IPv6 being deployed in existing IPv4 networks?

Agreed.  Randy is a well placed party :)

He made a very insightful and very correct statement here
http://ops.ietf.org/lists/v6ops/v6ops.2010/msg00978.html

Wireless is a weird edge, your normal recommendations and standard LAN
/ WAN logic does not really apply.

If you would like to take mobile networks into account, i strongly
recommend that you view Mobile section from the Google conference
https://sites.google.com/site/ipv6implementors/2010/agenda

What has been missing from our dialog is the fact that i have already
deployed IPv6-only edge (phones, not routers) nationwide in the USA
without a single national network MOP, ipv6 address on a production
router,  or router upgrade.  The radio networks is also oblivious to
IPv6 services.  All i had to do was deploy an anchor point in a lab
with an IPv6 ISP, tweak a few user profile settings, and the rest just
works.  Thank you 3GPP.

This is by virtue of the fact that the user's IP (v4 or v6) only
really lives on the handset and the mobility anchor (GGSN).  Once the
IP packet leaves the handset it is tunneled all around inside of IPv4
mobility tunnels until it reaches the Internet boarder of my network,
where it becomes a real IPv6 routed packet again.  So, literally, the
IPv6 packets are only ever exposed on the handset itself and on the
mobility anchor (GGSN) facing the Internet.  There are hundreds of
routers in between that are oblivious to the fact that IPv6 is
deployed because the mobility tunnels are all IPv4 on the outside...
and will stay that way ... since it is only the edge (phones, not
routers) that NEED IPv6 to grow *my* business.

Dual stack routers in my network don't matter, aside from my internet
peering.  That said, the handset can be provisioned as "non-unique
with my network" or BOGON IPv4, IPv6, or at 2x the *packet core* cost
with non-unique IPv4 addresses AND IPv6.

I hope this provides some clarity.  For mobile, it is not about the
"IP network".   IPv6 vs IPv4 vs DS is only about the handset and how
it instantiates its Internet communication on the mobility anchor.  It
is fundamentally different from a backbone ISP going to dual-stack
from IPv4, which is a smart path that all my upstream ISPs have done.

Regards,

Cameron