Global IPv6 policy question

[Geoff Huston]

At 12:13 AM 25/07/2006, Tim Chown wrote:
>On Fri, Jul 21, 2006 at 06:20:08PM +0200, Iljitsch van Beijnum wrote:
> > On 21-jul-2006, at 17:42, Kevin Loch wrote:
> >
> > >/12 seems to be rather large given the unlikely possibility of
> > >a /13 assignment being made any time soon.  The largest
> > >assignments being made so far have been /19, and only
> > >two ever.
> >
> > >A year's supply of the largest expected assignment size would
> > >be a good target.  /16 should be more than enough.
> >
> > /12 makes for 512 blocks under 2000::/3 so it doesn't seem too
> > excessive, even if a bit longer would also be sufficient.


Some folk on this list may be regular attendees of the open address 
policy meetings hosted by the Regional Internet Registries, and for 
these folk this topic will all sound very familiar, no doubt. In 
formulating this address policy proposal within these forums, this 
topic was considered in some detail. Rather than rehash all the 
material here (for which some of you may be grateful to be spared the 
agony of such a rerun!) those who are interested may find it useful 
to trawl back through the sessions of past RIR meetings where this 
topic has been considered.  In this context, I presented some work 
looking at simulations of registry activity in a V6 registry and 
looked at the impact of various policies in terms of the ability of 
the registry to support coherent (non-fragmented) allocation windows 
for ISPs and LIRs across the full sequence of initial and subsequent 
(window expansion) allocations. An unpublished draft on this topic 
can be found at 
http://www.potaroo.net/drafts/draft-huston-ip6-allocation-unit-00.html 
The aim of this exercise was to avoid the non-terminating qualitative 
discussion of "/x looks better than /y as an allocation unit for me" 
and instead look at the outcomes of choosing a particular value in 
quantitative terms of the ability for the allocation registry to 
avoid inherent address block fragmentation through discontinuous 
allocations. This probability of fragmentation of downstream 
allocations is a function of the allocation unit size, particularly 
if the allocation algorithm is based on the sparse rate-modified 
allocation algorithm. In any case I would urge those who are 
interested in this topic, who have not already done so, to spend some 
time looking back at the previous discussions of this policy proposal 
in the online archives of the open address policy sessions.