[arin-ppml] ARIN-prop-127: Shared Transition Space for IPv4 Address Extension

[Frank Bulk]

Comments in-line.

-----Original Message-----
From: Jack Bates [mailto:jbates at brightok.net] 
Sent: Saturday, January 22, 2011 1:39 AM
To: frnkblk at iname.com
Cc: arin-ppml at arin.net
Subject: Re: [arin-ppml] ARIN-prop-127: Shared Transition Space for IPv4
Address Extension

On 1/22/2011 12:06 AM, Frank Bulk wrote:
> Yes, eyeball networks benefit from this policy proposal more than content
> providers.  I don't believe that eyeball networks will leverage NAT444 to
> sell "extra" IPv4 -- the breakage that accompanies NAT444 is too painful.
> And if NAT444 was so successful, what market value would that IPv4 address
> space have?  It would be cheaper for those supposed buyers to implement a
> CGN.

1) Is it your opinion that the eyeball networks deserve to benefit from 
this proposal and eat up a /10 which could be utilized by content 
providers who don't get near the transition capabilities as eyeball 
networks?

FB> No, the eyeball networks don't deserve it, but I'm more interested in
moving the Internet forward through this crunch and the segment that seems
to be in the greatest need is the eyeball networks.  The content providers
have (fortunately) been able to scale using load-balancers and anycast.  If
you have a proposal for ARIN that would assist the content providers in
growing their business in the face of IPv4 depletion, please send one in.  

2) Given that eyeball networks for support reasons will likely push 
their cheapest plans into NAT444 quickly (so as to support a uniform 
topology), what do you believe they will do with all this extra space 
gained? Others in the thread have mentioned v4 value projected to be $40 
per? Many eyeball networks also sideline these days with content 
services as well. Will they utilize this massive block of free IPv4 they 
recover to push out the content only providers?

FB> I'm not sure if service providers will apply NAT444 to their "cheapest
plan" customers first.  The extra support required for NAT444 customers
would only further reduce their margins on this customer segment.  Eyeball
networks won't consider any IPv4 they've freed up as 'free'.  They had to
buy equipment to do the NAT444 and field extra support calls, while their
competitors, who had sufficient space, don't do any CGN at all and boast
about it.  But the future could prove me wrong.

3) While NAT444 is painful and there is hope that many things will 
utilize v6, what percentage of services do you see remaining v4, where 
NAT444 ceases to be painful? For example, http works rather well with 
NAT444, so the pain threshold is much higher than say... skype/p2p. I 
expect skype and other p2p programs (many already do) to support ipv6 
quickly to protect their business model from dealing with NAT444, while 
many services which don't have breakage from NAT have no driving force 
to push them quicker to IPv6. The same goes for the eyeball customers. 
They are likely to quit using home routers if it means gaining IPv6 
connectivity to make their xbox/skype/wow patch work (or upgrade to an 
IPv6 capable device). This will v6 enable them, and then they'll just 
utilize v4 for the services which haven't migrated as quickly as NAT444 
doesn't break them (or the break, such as geolocation, ip specific 
rules, etc are acceptable by both parties).

FB> Many good points here.  We can't know what impact NAT444 will have on
extending the long tail of IPv4 content because IPv4 eyeballs remain.  But
there will be eyeballs, eventually, that are IPv6 only, so anyone who keeps
their content IPv4-only will miss some eyeballs.  

At this stage of the game, I'm sure it's too late to correct policies. 
ARIN should have had policies in place already to help protect the 
content providers over the eyeball networks, due to the unbalanced 
transition tools available. Perhaps this policy is the sacrifice (we'll 
give you a /10, if you'll just be quiet and not eat up more space; 
accept all that extra space you'll reclaim on NAT444 and not even worry 
about the 1-10% needed to do NAT444). Or perhaps implementing this 
policy won't stop the flood of requests, and it will just sacrifice 
another /10 to eyeballs that content providers could have used. I didn't 
see anything in the policy that mandated all justifications must show 
NAT444 using this /10 for the customer facing side. Without such 
wording, a provider could still utilize NAT444 to justify more space, 
and then turn around and utilize the /10; giving themselves a better 
buffer on runout (or more of those so call $40 addresses).

FB> Yes, there will be service providers who do just as you described, but
we shouldn't penalize all those who can't currently justify additional
requests, anticipate future growth, acknowledge that NAT444 is the
transition technology they need, and need a path moving forward.

For the record. I'm an eyeball network, and content I provide falls 
under standard eyeball network services (and the bare minimum at that). 
I don't suffer from the issues many content providers will face, but I 
will see the effects when they have issues obtaining IPv4 space while my 
network and many other eyeball networks free up IPv4 space through 
NAT444 and are still loathe to give up claim to the space (or bank by 
selling it to the content networks at premium rates). I'm all for my own 
network paying the tax of sacrificing existing space for utilization in 
NAT444.

FB> I hope that you can avoid NAT444.  I'm hoping to, too.  In the words of
Randy Bush, I encourage my competitors to pursue NAT444.

Jack