[arin-ppml] ARIN-prop-127: Shared Transition Space for IPv4Address Extension

[Owen DeLong]

On Jan 21, 2011, at 1:53 PM, George Bonser wrote:

> 
>> From: Chris Donley 
>> Sent: Friday, January 21, 2011 1:49 PM
>> To: George Bonser; Chris Grundemann
>> Cc: arin-ppml at arin.net
>> Subject: RE: [arin-ppml] ARIN-prop-127: Shared Transition Space for
>> IPv4Address Extension
>> 
>> 
>> Still, my major source of discomfort is enabling v4 forever.
>> [CD] I think the market will take care of that.  NAT444 is going to
>> suck for all but the most basic services.  It will offer a degraded
>> quality of experience for video streaming, gaming, voice, etc. -
>> services customers want to use.  IPv6 will offer a better quality of
>> experience through bypassing the NAT.  Customers will get the message
>> and start using IPv6 as they replace legacy devices. As Owen said,
>> NAT444 is a great business case for IPv6. I don't see any way around
>> NAT444, though.  It's the only IPv4 extension technology that's
> readily
>> deployable and doesn't require a new home gateway.
>> 
>> Chris
> 
> Why is it going to suck?  We are actually doing it now, pretty much.  A
> computer in an office opens a connection to a content provider.  That
> RFC1918 packet gets NATed to a global IP somewhere.  That packet then
> hits a load balancer at the content provider where it is again
> translated and directed to a machine in 1918 space again.
> 
Yeah, well... Try this experiment to decide how much your customers
would like it:

Xbox -> RFC-1918 network -> NAT Box (with UPnP) -> different RFC-1918 network -> NAT Box (without UPnP) -> Internet

Now, try to host an Xbox live game for your friends to join.

Try this with World of Warcraft, including downloading an update
or two, and a couple of 25-man raids.

Try AIM, Yahoo Instant Messenger,
Skype (be sure to test each of these messengers with audio
and video and file exchanges).

Try to stream a Netflix video over IPv4 in that environment.

Try to use your Vonage phone (or any of several other VOIP
telephony services).

> For all practical purposes most of the traffic in the v4 Internet today
> is already NAT444
> 
Not in the meaningful and harmful way that it will be under what we
are calling NAT444. No.

> 
> In and of itself it isn't going to break anything that doesn't already
> break today unless someone tries to use an underpowered box to do the
> NAT.
> 
This simply is NOT true. See above.

Today, what you have at worst is:

customer <-> NAT (controlled by customer) <-> Internet <-> NAT (controlled by content provider) <-> Content

What people mean when they say NAT444 is

cust<->NAT (Ctl by Cust)<->ProviderNet <->NAT (ctl by Provider)<->NAT (Ctl by CP) <-> Content

That NAT not controlled by either the cust or the CP in the middle will
actually break lots of things and because neither side can intervene
to control the behavior of the NAT that is breaking things or gain
additional hints about the translation of state on either side of the
box, the number of workarounds available is much more limited.



Owen