[arin-ppml] Do we have a fundamental conflict in theregionalgoalsfor number resource management?

[Mike Burns]

Hi David,

Thank you for your feedback.

If disaggregation is going to be a problem, I suppose we will see that first 
in APNIC because of the lack of justification requirements.
I hope we can rely on Geoff Huston to keep us informed.

But if performance degredation due to disaggregation is a real issue, 
network operators will refuse, as they do today, to route blocks too small.
And so anybody buying a small block will not be paying anything like top 
dollar.

In fact, the boundaries on routable size are strictly in the hands of the 
network operators anyway. I do believe market and technological forces will 
work on the network community that will have the eventual effect of 
decreasing routable block size. Every transfer I've ever seen has been a /24 
or larger, but I think we know that /25 advertisements will come eventually.

If the delta between the price of small blocks and larger blocks is very 
high, somebody will come in to sweep up the dregs into routable sizes.

I don't know how it will develop. Maybe one /8 will become known for small 
blocks, and an aggregator may work only in this space. Maybe there are some 
economies of scale which increase the success of marginal aggregations  with 
very large pools. Maybe the pools will be virtual pools of people wishing to 
deaggregate from their existing routable sizes, except they won't pull the 
trigger until a suitable pairing is made. Maybe people will shop for 
discrete blocks that they know they can aggregate themselves, or keen-eyed 
traders will do that work. Maybe, as you said, it's not feasible.

On a related topic, from the list of the (10) 8.3 transactions to date, it 
doesn't look like the needs analysis requirements in place for those 
transactions led to any aggregation.

Rather than 10 single aggregates, there were a substantially higher number 
of netblocks involved. So I'm really not all that sure that removing the 
needs requirement would work any more rapidly towards table growth than 
current policies.

Thanks for sharing your experience with the 5 /16s. That is directly 
relevant, and maybe there is no future in the aggregation business.

Though I still hope that one day somebody will put that last phrase on a 
plaque, along with "Heavier than air flight is impossible", and mount it in 
the lobby of their aggregation business!

And at the risk of being tiresome, there are still the problems with current 
needs policy that I elucidated earlier, which problems I believe outweigh 
the risks posed to the size of the BGP table.

I'm going to find and reread an article by Geoff Huston which related BGP 
growth rates and ASN growth rates to see if anything can be added to this 
discussion. Seems to me he found some mysterious bounding effect which 
caused BGP to grow less slowly than the steady ASN growth, but I could be 
wrong on that.

It really helps to have the information that you and Tony Li and Bill Herrin 
provided.

I think with a couple of days more discussion that I will be ready to begin 
generating a proposal, and I thank the members who have expressed support 
for the idea of removing all needs analyses from transfers.


Regards,
Mike



----- Original Message ----- 
From: "David Farmer" <farmer at umn.edu>
To: "Mike Burns" <mike at nationwideinc.com>
Cc: "Public Policy Mailing List" <ppml at arin.net>; "David Farmer" 
<farmer at umn.edu>
Sent: Friday, April 29, 2011 8:15 PM
Subject: Re: [arin-ppml] Do we have a fundamental conflict in 
theregionalgoalsfor number resource management?


> On 4/29/11 12:58 CDT, Mike Burns wrote:
>> But I can see the development of aggregation entities with pools of
>> addresses who make money on aggregation services. As in, I have 4 /25
>> networks. Can I turn them into you in exchange for a single /24? And the
>> aggregator transfers a separate full /24 puts the /25s in inventory
>> until he can purchase the other half of the /24s which would allow him
>> to aggregate them from /25s back up to /24s, or until such time that the
>> network operator community decides to generally accept /25 
>> advertisements.
>
> Have you run the numbers on the churn rate you would need to have any 
> probability getting aggregate-able blocks to come available in an 
> acceptable return-on-investment timetable for a commercial entity to be 
> willing to operate that kind of business model?
>
> I work for an organization that received what will be its last allocation 
> of IPv4 addresses almost 19 years ago, it is only two days away.  We have 
> been actively recycling out of our 5 /16 blocks for more than 10 years. 
> I can tell you, even in a much smaller address space and with the ability 
> to actually sometimes force readdressing, re-aggregating blocks is 
> extremely hard to do.
>
> I find it extremely hard to believe there is a viable business model for a 
> commercial entity to make this happen at Internet scale, especially with 
> the realistic churn rates most people expect to see.  There isn't going to 
> be enough additional value created by aggregating blocks to justify the 
> loss of potential revenue from holding on to the blocks too long.  Yes, it 
> will happen, but not reliably enough to make a business model out of it.
>
> Additionally, the costs of disaggregation are too dispersed to find a way 
> to capture any of the potential Internet-wide cost savings from 
> aggregating blocks.  So, I don't see how the economics of a business model 
> based on aggregation of blocks would work.  Maybe, I'm missing something. 
> :)
>
>
> -- 
> ===============================================
> David Farmer               Email:farmer at umn.edu
> Networking & Telecommunication Services
> Office of Information Technology
> University of Minnesota 2218 University Ave SE     Phone: 612-626-0815
> Minneapolis, MN 55414-3029   Cell: 612-812-9952
> ===============================================