On Jan 31, 2012, at 12:32 pm, Andy Linton wrote:

[…]

2. Summary of the current problem
---------------------------------

The current IPv6 address allocation and assignment policy
(apnic-089-v010) states that

  5.2.3 Larger initial allocations
  Initial allocations larger than /32 may be justified if:
  a. The organization provides comprehensive documentation
     of planned IPv6 infrastructure which would require a
     larger allocation; or
  b. The organization provides comprehensive documentation
     of all of the following:
     o its existing IPv4 infrastructure and customer base,
     o its intention to provide its existing IPv4 services
       via IPv6, and
     o its intention to move some of its existing IPv4
       customers to IPv6 within two years.
  In either case, an allocation will be made which fulfills
  the calculated address requirement, in accordance with
  the HD-Ratio based utilization policy.

Large networks are facing challenges deploying IPv6 networks. The
current slow start policy is to allocate a /32 and then reduce the bit
mask one bit at a time on subsequent allocations (i.e. /31, /30, /29
etc.).

It would be helpful if the authors of the proposal could expand on this explanation of the issue being solved. I do not understand why the slow start policy is relevant to large scale IPv6 deployments on existing IPv4 networks. I would have expected that in most cases APNIC, or the relevant NIR, would already have information about the network on which IPv6 is being deployed and would just need details of the new addressing plan, as per 5.2.3.b.

Why is slow start an issue here? What have I misunderstood?

The "slow start policy" itself is not an issue. The issue is that the reserved IPv6 address pool is only considered for TWO years.  If we can expend the time window to five years or even ten years, there should be no problem for the slow start policy.

This approach is designed to maximise global routing aggregation,
however, it causes fragmentation and complexity in the internal routing
configuration of very large networks. This is particularly a problem in
large networks with many POPs growing at different rates.

Also, the IPv6 Address Allocation and Assignment Policy (Section 5.2.3
Larger initial allocations) does not take into account long-term future
growth.

What is the scale of mismatch between large scale IPv6 deployments on existing IPv4 networks and anticipated "long-term future growth"? What is "long-term" here? 3.7 already says that LIRs should not have "to go back to RIRs for additional space too frequently", so plans for relatively short periods, like four or five years shouldn't be a problem.

I am obviously missing something. What is it?

Thanks,

Leo Vegoda
*              sig-policy:  APNIC SIG on resource management policy           *
_______________________________________________
sig-policy mailing list
sig-policy@lists.apnic.net
http://mailman.apnic.net/mailman/listinfo/sig-policy