[arin-ppml] IPv4 runout happens when the first ISP deploys CGN+IPv6*

[Lee Howard]

----- Original Message ----

> From: Matthew Kaufman <matthew at matthew.at>
> To: Lee Howard <spiffnolee at yahoo.com>
> Cc: Jason Schiller <schiller at uu.net>; Kevin Kargel <kkargel at polartel.com>; 
>"arin-ppml at arin.net" <arin-ppml at arin.net>
> Sent: Thu, December 30, 2010 3:44:09 PM
> Subject: Re: [arin-ppml] IPv4 runout happens when the first ISP deploys 
>CGN+IPv6*
> 
> On 12/30/2010 12:31 PM, Lee Howard wrote:
> >
> > Content does not only  live on "servers."
> >
> > Once ISPs can't get new IPv4 addresses, they  may pay for transfers for
> > hosted servers, but access customers will get  CGN and hopefully IPv6.
> > If one of those customers is a p2p seeder, or is  your online gaming buddy
> > (through any game console), he is unreachable  via IPv4.
> Only for peer-to-peer protocols that can't traverse a  CGN.

If the *seeder* is on a home gateway, behind a CGN, there's no workaround.
It has an rfc1918 address, and maybe it's used ICE/TURN/STUN or uPNP
or other magic to get through the local gateway, so it tells the tracker it's
"outside" address, which is the address of the home gateway.  But the
tracker is outside the CGN, so it provides the ISP's private address to the
client.  I've only lab tested this, but I've never gotten it to work.

This is also true of gaming consoles.  The online server that brokers the
connection uses the source address of each console, which may be the
outside of the CGN.  But the CGN doesn't have a mapping for client
to client, only client to server, so the game setup fails.  I've lab tested
this too, and it fails slightly differently for different game consoles.

See http://tools.ietf.org/html/draft-donley-nat444-impacts-01

My points are:
* NOT that we need better NAT traversal, but that we need better IPv6
support.
* IPv4 only "continues to work" in some contexts.  

Lee