[arin-ppml] IPv6 End-User Initial Assignment Policy (or: Please don't me make do ULA + NAT66)

Adam Thompson athompso at athompso.net
Wed Feb 18 18:14:32 EST 2015

On 2015-02-18 04:47 PM, james machado wrote:
> So we argue for a /48 for each home user site but we toss out that
> argument when it comes to a smaller business with multiple sites?
> I applaud the intent but think it is too short sighted William. It
> should take no more routing slots for an aggregated /40 or /44 than
> for a /48 and the /40 or /44 are in line with the v6 paradigm that has
> been fronted on this list and others.

Given the high demand for IPv6 allocations (*ahem*), the minimum 
requirement could easily be eliminated altogether, with relatively 
little impact AT THIS TIME as far as I can see.
I think having to justify anything larger than a /48 is reasonable, 
otherwise we probably would get people asking for /8s just to camp on them.

I fully agree with the OP's point that not being able to get PI space 
represents a business risk, and that business (i.e. 2nd-order financial) 
risk does not scale in any way an engineer or scientist would expect.

I have one regional ISP customer who could probably renumber in two 
weeks with relatively little impact, whereas I have one large (Fortune 
500-sized, but private) multinational enterprise customer who could be 
in danger of bankruptcy if they lost one single IPv4 /24.

My point here is that there *is* business risk, sometimes enormous risk, 
tied to using non-PI space.  As technologists we find this next bit 
uncomfortable: there's no formula to predict the level of risk.  This 
makes writing policy to address it very difficult, but writing policy 
that ignores it is worse.

I've been using the multi-homed clause for all the non-SP customers I 
work with, so far, since none of them have enough public IP resources to 
warrant a direct allocation otherwise. Even in the mid-western Canada 
telecom wasteland, it's almost always possible to get two connections... 
even if one is a tiny local ISP who doesn't even run BGP or IPv6 
themselves, that still satisfies (IIRC) the letter of the NRPM.  
(Besides, then there's a reason for that tiny local ISP to improve rapidly.)

This is, unsurprisingly, why I was very unhappy to see the multi-homed 
clause vanish from the IPv4 section.

-Adam Thompson
  athompso at athompso.net
  +1 (204) 291-7950 - cell
  +1 (204) 489-6515 - fax

More information about the ARIN-PPML mailing list