[ppml] Policy Proposal: IPv6 Assignment Guidelines

[David Conrad]

On Aug 18, 2007, at 9:28 AM, Paul Vixie wrote:
>> However, people are finding interesting ways to be irrational.
> like using a /64 for a subnet?

2^64 subnets is still an insanely large number.  While it is ...  
questionable (particularly given how most large scale enterprise IT  
folk get twitchy when you say "stateless auto-configuration"), I  
don't see this as a real problem.  It just turns the address space  
from 2^128 to 2^64.

> or like PI /48's to recreate the IPv4 swamp?

No.  I was speaking more of allocating /20s, /19s, and shorter,  
particularly given the liberalization of PI allocation policies.   
People seem to forget that there are the same number of /19s (etc) in  
IPv6 as there are in IPv4.

You want to conserve address space?  Never _EVER_ allocate longer  
than a /32.  If a requester has demonstrated they have consumed the / 
32, grow it to a /31.  If they have demonstrated they've consumed  
the /31, grow it to a /30.  Etc.  Never reserve space or allocate  
sequentially, simply use sparse allocation with bisection (that is,  
after all, why each RIR received a /12).

Speaking of that, why is ARIN still allocating IPv6 sequentially?

Regards,
-drc