[ppml] dual-stack (was Re: Expand timeframe of Additional Requests )

[Stephen Sprunk]

Thus spake "Paul Vixie" <paul at vix.com>
>> What's wrong with NAT-PT?
>
> see http://tools.ietf.org/html/rfc4966, i guess.

I'll save eveyone else some time reading: NAT is bad therefore NAT-PT is 
bad.

>> My ideal CPE box would provide both v4 and v6 downstream.  If the
>> upstream was v6-only, it would get the inside prefix via DHCP PD and
>> do NAT-PT for v4 traffic.  If the upstream was v4-only, it would do 6to4
>> for v6 traffic.  If both were available upstream, it'd do DHCP PD and
>> native routing for v6, and normal NAT and DHCP for v4.
>
> you'll need to present a case for this to ietf/v6ops, and offer a solution
> that lacks NAT-PT's statefulness and modal complexity.

The IETF isn't interested in any transition model that doesn't require 
dual-stacking every host on the Internet before the first v6-only node 
appears.  This is a fatal flaw that will ensure every model they come up 
with fails.  Oh, and there's a moratorium on "new" transition models as 
well, so even if one had a great idea, it would be ignored on principle.

NAT exists, despite IETF bleating about its evils, and these days virtually 
every host on the Internet is behind one.  NAT-PT extends that evil, yes, 
but it gets us out of the chicken-and-egg problem we have today where nobody 
is willing to pay the cost of deploying v6 because nobody else has deployed 
v6.  NAT-PT would provide a critical mass of hosts that _appear_ to be 
dual-stacked, giving other sites motivation to deploy v6 themselves and 
making it possible for v6-only networks and hosts to be deployed.

> i agree with the general goal that a CPE could let V4-only devices 
> continue
> to connect but not depend on a V4 ISP infrastructure other than at the 
> "far
> edge".

My model stil assumes that two v4-only hosts would speak native v4 to each 
other, though likely through NAT devices -- which is already common.  An 
automatic v4-over-v6 tunneling mechanism is something we won't need for many 
years to come.

>> The best case I see is that ISPs will deploy native v6 and then stuff all
>> their customers (well, at least residential) behind a giant v4 NAT box 
>> per
>> POP, handing out addresses in 10/8 via DHCP.  That'd allow them to drop
>> their v4 allocations, yet still provide "pure" access on v6, motivating
>> customers to enable it.
>
> would that work?  (is there a proof of concept running anyplace?)

I can't see any reason it wouldn't work, but I also don't see the motivation 
for anyone to try today, since there's no native (or even NAT-PT'd) v6 hosts 
for their customers to talk to.  I'm not aware of any proof of concept.

S

Stephen Sprunk      "Those people who think they know everything
CCIE #3723         are a great annoyance to those of us who do."
K5SSS                                             --Isaac Asimov