[ppml] Policy Proposal: Global Policy for the Allocation of the Remaining IPv4 Address Space

[Brian Dickson]

Leo Vegoda wrote:
>> Changing the size of CIDR blocks given to RIRs, at some point, would
>> increase the number
>> of blocks available, thus ensuring that fair allocations continue to
>> occur.
>>
>> E.g. when there are 6 of the /8 ranges left, start allocating /10's, of
>> which there would be 24.
>> And when there are 6 of the /10's left, start allocating /12's, etc.
>
> Do you want to change the unit size or the maximum amount of space
> that can be allocated? The current global policy allows the RIRs to
> receive all the space they need and changing the way it is measured
> from /8s to /10s doesn't change anything other than add extra lines to
> the IPv4 registry.
Unit size.
Clearly this method won't support maximum space allocations that exceed
available space - but no method can.

Maximum space would likely be a function of number of blocks left at
current unit size, and/or some kind of "oversubscription" rules that
relate to how many additional blocks (of decreasing size) get assigned
to satisfy requests for blocks. The sequence of block sizes would be
strictly deterministic. 18 /10's, 18 /12's, 18 /14's, 18 /16's, etc.,
until the plug is pulled via the last 6 blocks being
assigned without subdivision.

If collectively the RIRs are about to hit the exhaustion of IPv4 space
"wall", then the presumption is more frequent and smaller allocations
make it possible to continue to serve up *something*, rather than having
no more space.

It's a bit of Zeno's paradox, put to practice. It is, however,
fundamentally fair, since there continue to be blocks available, albeit
decreasing in size and increasing in frequency of assignment.

At some point, there will need to be a final assignment of blocks to
RIRs, but I don't think that should happen until the block sizes are
substantially smaller than /8, if fairness is important.

(And yes, I believe it is, in the allocation of the last remaining
portions of a finite resource.)

Brian