[arin-ppml] Routing table growth, was Re: IPv4 is depleted today

[Tom Vest]

On Sep 3, 2008, at 11:57 AM, Iljitsch van Beijnum wrote:

> On 3 sep 2008, at 17:32, Tom Vest wrote:
>
>>> It's not the cost: it's the benefits. If it costs $50 to implement  
>>> v4
>>> and you can reach 500 sites out of the Alexa top 500, that's $0.10  
>>> per
>>> destination. If v6 is only $40 but you can only reach the 0.4% of  
>>> the
>>> Alexa top 500 sites that have IPv6 that's $5 per destination.
>
>> The question is, what mechanisms will contribute to increasing that  
>> benefit number?
>
> I don't think being able to reach IPv6 destinations in the immediate  
> future will drive IPv6 adoption.


I agree -- which is what makes me wonder how, post-runout, when  
presumably all future new entrants will be either pure IPv6 or  
absolutely dependent on transfer mechanisms, we're going to be able to  
continue looking like an open industry to both inside and outside  
critics. The way the political winds are blowing, that could easily  
become a real concern in the near future, even in the US.


> The driver will be the lack of IPv4 space for ISPs. They will then  
> have to choose a mechanism to get content (which will presumably  
> still be predominately on v4) to their users. One way is to do NAT  
> in the ISP network. Another is NAT-PT (well, probably NAT64,  
> (hopefully) its successor) and yet another "dual stack lite".
>
> I'll assume the first familiar, the second is IPv6 clients talking  
> to IPv4 servers through a v6-v4 translator + NAT and the third IPv4  
> clients talking through an IPv6 cloud and an IPv4 NAT to an IPv4  
> server.
>
> The last two both have the advantage that an ISP can support many  
> customers behind a big fat NAT box without having to do routing/ 
> addressing trickery and there's just the one NAT box. With the first  
> solution internal routing and management becomes a nightmare and if  
> there are additional NATs peer-to-peer becomes REALLY hard.


As the risk of erring on the cynical side, I suspect that such  
solutions will be deployable just as easily in non-transparent/ 
asymmetrical as in symmetric/transparent (i.e., transparently  
accessible to externally-initiated connections) mode. Given how the  
endless wrangling between content-oriented and access-oriented  
providers has heated up lately, it's easy to imagine how normal good  
hygiene, complexity aversions, et al. might not be sufficient to  
assure that internal fixes have the same kind of beneficial external/ 
inter-domain side-effects that we might otherwise expect.


> So NAT64 and/or dual stack light, along with futureproofness, may be  
> what gets IPv6 in the door at ISPs. Then, when the number of IPv6  
> eyeballs is sufficiently large, it becomes interesting for content  
> providers to talk to those people directly over v6, this helps them  
> track people more easily, among other things.
>
>> 1. Incumbent IPv4-based operators altruistically make their own  
>> content and services transparently accessible to remote IPv6-only  
>> networks,
>
> 0.4% of the Alexa top 500...

I understand the point you're trying to make here, but sometimes the  
elephants have ideas their own. Unless we introduce some contradictory  
altruism assumptions in here, it's not clear why the elephants -- most  
of whom have big interests in both access and content -- will be  
inclined to automatically empower the legions of mice, given the  
option of not empowering them.

> Note that once we're at about 10% or so IPv6 use, it really doesn't  
> matter all that much if it grows beyond that, until we reach 50/50  
> and then, 90%. At that point, people will start turning off IPv4.  
> Between 10% and 90% we'll have a de facto dual stack internet.


Agreed, my concerns are all about the path between those two  
milestones -- whether it's relatively straight or twisty turny with  
cycles, and how we're going to know where we are once we get in the  
middle and there's no turning back.


>> 2. Topologically adjacent IPv6-only networks grow and proliferate,  
>> driven solely by IPv6-based demand for IPv6-based content and  
>> services,
>
> Peer-to-peer. I already have IPv6 bittorrent peers regularly these  
> days.
>
>> 3. IPv4 transfer prices and/or outsourced v4/v6 translation  
>> services remain low enough so that otherwise IPv6-only networks can  
>> exchange traffic with each other remotely without being  
>> substantially / differentially disadvantaged relative to IPv4-based  
>> incumbents.
>
> I don't understand.

> On Sep 3, 2008, at 10:42 AM, Iljitsch van Beijnum wrote:

>> It's not the cost: it's the benefits. If it costs $50 to implement v4
>> and you can reach 500 sites out of the Alexa top 500, that's $0.10  
>> per
>> destination. If v6 is only $40 but you can only reach the 0.4% of the
>> Alexa top 500 sites that have IPv6 that's $5 per destination.



I was just suggesting that post-runout, aspiring new entrants will  
have to use very different denominator(s) to calculate the benefits of  
any/all IPv4 reachability. Maybe being a dedicated customer will  
remain just as competitive as it is now, but I don't think that will  
fact make for a very compelling defense if the price of becoming a new  
PI self-provider or entering the market as a new IPv4 routing services  
provider goes astronomical.

I just can't see what will prevent that outcome, short of the sort of  
unrealistic assumptions of altruism or coordinated action plans that  
have been ruled out of bounds.

TV