[arin-ppml] Update on 2009-3: Global Policy for the Allocationof IPv4 Blocks to RIRs

[Scott Leibrand]

Stephen Sprunk wrote:
> Scott Leibrand wrote:
>   
>> I'm not sure it makes sense to re-write the global policy any more
>> than we've already done, but let me see if I can explain the
>> possibilities.  As a background point, remember that global policies
>> are supposed to define the RIRs' interactions with IANA, while local
>> policies define the interactions between RIRs and the organizations in
>> their region.
>>
>> - RIRs develop their own local policies and procedures  as to
>> whether/how they want to reclaim IPv4 space.  If nothing is returned,
>> do nothing here.
>> - RIRs develop their own local strategies and policies for deciding
>> what space goes back to IANA.  If nothing is designated, do nothing
>> (until someone complains, and a policy/procedure to return space is
>> adopted).
>> - RIRs return any such designated space to IANA at quarterly intervals
>> in aggregated blocks of /24.
>> - IANA creates a returned address space pool.
>> - When the IANA free pool is exhausted, and RIRs need more space, they
>> can request space from the returned pool.
>> - IANA gives out space to requesting RIRs, as defined in this global
>> policy.
>>     
>
> This makes sense, except that it directly leads to fragmentation across
> RIRs.
>
> I oppose any policy that allows an RIR to return space to IANA in any
> smaller unit than IANA originally allocated that particular space to an
> RIR or directly assigned it to an end-user org (e.g. class B/C legacy
> blocks in the "Various Registries" /8s) or does not require IANA to, to
> the extent technically feasible, aggregate such returned blocks before
> reallocating them to the most logical RIR.
>
> For instance, imagine x.y.0.0/16 is currently APNIC and x.z/16 is
> currently RIPE, and the two prefixes could by aggregated into a single
> /15; if the former is returned, it should be offered to RIPE first, and
> if the latter is returned, it should be offered to APNIC.  Neither
> should be allowed to return a fraction of their block but should instead
> reuse the remainder -- or encourage the remaining users to renumber out
> so that the entire block can be returned.
>
> Frankly, though, I don't see much point in this policy.  When we hit the
> wall, everyone is going to be wanting more space and there will be none
> in IANA's return pool.  When we finally come to our senses and switch to
> IPv6, there will be loads of returns that nobody will want and that the
> RIRs might as well hold on to in case of future stupidity in their
> region.  Why make extra work for IANA just because one region might be
> dumber than the others?  It'd be one thing if we were moving entire /8s,
> but mere /24s?  That's nuts.

This is part of the reason for making the designation of space to return 
to IANA a matter of local policy.  All five RIRs need not agree in 
advance on how big a block must be to be "of global significance" and 
worth returning.  And in fact, that size is likely to shift over time.  
Instead, that can be defined in local policy at each RIR.

I think everyone, including the authors of the global policy, 
acknowledges that the actual impact may turn out to be small.  However, 
the risks of not setting up a structure for redistributing reclaimed 
space, in the event it is needed, are much higher than the risk of 
setting up such a structure, and then not using it (much).

-Scott