[arin-ppml] Preemptive IPv6 assignment

Andrew Koch andrew.koch at gawul.net
Fri Oct 8 01:38:42 EDT 2010

On Thu, Oct 7, 2010 at 18:31, Matthew Kaufman <matthew at matthew.at> wrote:
>  On 10/7/2010 1:53 PM, Andrew Koch wrote:
>> While I like the idea of making it easy as possible to get IPv6 space,
>> there needs to be some planning and responsibility behind getting this
>> space.
> Why? There's lots of it.

Sure there may be lots of it, but what happens when you pre-assign a
/32, but the org needs a /26?  They start using their pre-assigned
addresses as you have directed for them to do, but the soon run out
and would be very painful for them to renumber into the appropriate
sized block.  Having to allocate/assign multiple times may result in
an inability to aggregate.

> I have an even easier idea: Ignore what's visible in BGP and simply direct
> IANA to allocate a portion of IPv6 space such that:
> <prefix>:<ASN>/<size>.
> A few choices might be:
>    A) prefix of 16 bits, full 32-bit ASN, size=48 (this is the obvious one,
> but then lots of entities need to come back for more, as they really are
> ISPs to more than just customers who'll take a /56 or /64... but it is a
> fairly conservative "experiment" to only use a single 16-bit prefix)
>   B) prefix of 3 or 4 bits, 28 or 29 bits of ASN (hoping that we never
> allocate that many 32-bit ASNs), size=32 (this is the one that means that
> almost nobody ever needs to come back for more)
> Nice compromises include:
>    C) prefix of 4 bits, full 32 bits of ASN, size=36 (I like this as much or
> better than choice B, but one could argue that it is a lot of v6 space to
> use)
>    D) prefix of 8 bits, full 32 bits of ASN, size=40 (not bad, and doesn't
> use up nearly as much IPv6 space if it turns out to be a bad idea)
> Then every ASN, legacy* and new, in all regions, immediately gets a
> reasonable amount of IPv6 space without any additional paperwork or database
> entries.
> The only problem created by this (especially if "size" is up near or at /32)
> is a financial one for the registries... but they're going to have that
> problem with nobody coming back for additional v6 assignments anyway.
> Matthew Kaufman
> * I will observe here that there are lots of entities who have legacy IPv4
> space and legacy AS numbers who currently aren't party to a LRSA and don't
> pay anyone anything. By definition, they were early adopters of IPv4. One
> might suppose that such folks might also be/have been early adopters of
> IPv6... but they are in a particular bind if they want to keep their status
> with regard to (in this region) ARIN at the same time as they try to be an
> early IPv6 adopter.

I also believe this creates more PI space that may not have been
otherwise needed.  This again leads to further de-aggregation and more
burned slots in expensive router memory.  So, besides the RIR possibly
loosing out on money, all routers that are in the DFZ are impacted as


More information about the ARIN-PPML mailing list