[arin-ppml] Draft Policy ARIN-2013-3: Tiny IPv6 Allocations for ISPs

Owen DeLong owen at delong.com
Thu Mar 28 01:13:47 EDT 2013


On Mar 27, 2013, at 6:34 PM, David Farmer <farmer at umn.edu> wrote:

> On 3/27/13 19:40 , Jimmy Hess wrote:
>> On 3/27/13, John Curran <jcurran at arin.net> wrote:
>>> On Mar 27, 2013, at 7:52 PM, Owen DeLong <owen at delong.com> wrote:
>>>   How many customers does an typical xx-small ISP have today?
>>>   (xx-small being those ISPs who hold a /22 of IPv4 space)
>> 
>> Well, logically they could have up to  approximately1000 customers,
>> assuming no NAT
>> and an average of one /32 per customer.
>> 
>> The IPv6 equivalent of  1000 /48s     =     a  /38
> 
> Well technically 1024, but with an 80% rule that is 819 customers. However, with a residential ISP using customer pools and lets say one pool of /22 then 512 customers could justify more IPv4, I think.  But with /29 pools you would justify more address space with 80% of the /29 pools allocated or 103; and 50% or more usage in all pools, or 4 addresses; So that could be as little as 103 * 4 = 412 residential customers.  So the CPE for those residential customers are going to request what size blocks using DHCPv6-PD, probably /64s mostly today, but some might request /56s and hopefully some will request /48s. Obviously a /40 wouldn't provide enough /48s, but I'm not sure /48s are realistic.

Note, we don't really have an 80% rule in ISP IPv6… It's 75% at two levels, so the number is actually 768.

/48s are realistic. We certainly have the opportunity to make them non-realistic by discouraging development of gateways that support them by having policies and/or fees that prevent ISPs from giving them out.

> While on the other had, you would only need 103 /29 business customers to justify more IPv4 space.  And a /40 provides more than enough /48s to cover this business customer scenario.

You do realize that there are a number of ISPs that (quite legitimately) provide /29s to residential customers.

Owen





More information about the ARIN-PPML mailing list