On Thu, Mar 21, 2013 at 11:24 PM, Tore Anderson <tore@fud.no> wrote:
Hello Scott,
* Scott Leibrand
While I support the goals of 2013-03 in a post-IPv4-depletion world, we are currently in an uncomfortable intermediate state where IPv4 depletion has occurred in some regions (RIPE and APNIC) but not in others (ARIN, LACNIC, and AfriNIC). I think it will be wise to wait until IPv4 depletion to remove of (at least some) needs requirements in those regions that still have an IPv4 free pool.
Geoff Huston's http://www.potaroo.net/tools/ipv4/index.html projects that global IPv4 depletion won't happen until July 2020, when AfriNIC is set to deplete as the last of the five RIRs. That is a *long* time to keep around a bureaucracy that no longer serves any purpose.
I was not very clear. I was just making the (obvious and possibly irrelevant) point that LACNIC and AfriNIC won't want to do something like 2013-03 until their free pools are exhausted.
As a result of the fact of remaining RIR free pools, and the current policy and sentiment in the ARIN region that inter-RIR IPv4 transfers should occur only to organizations/regions that justify need for the addresses (to avoid a run on said free pool), I think it would be wise to do something like this:
- First, pass an inter-RIR transfer policy in the RIPE region that is compatible with both the APNIC and ARIN inter-RIR transfer policies (i.e. has some form of needs justification). - Second, make sure that RIPE's transfer policy serves all organizations in the RIPE region, including those who in the past got PI space from RIPE. - Third, relax the ARIN region's inter-RIR transfer policy such that after IPv4 depletion in the ARIN region, transfers are allowed to regions with policies like 2013-03. - Fourth, pass something along the lines of 2013-03 in the RIPE region (to take effect) after IPv4 depletion has occurred in the ARIN region. - Fifth, update policy in other regions as well to align policy with the needs of a post-IPv4-depletion world.
I'm more than happy to help drive #3 and eventually #5 in the ARIN region.
While I see and acknowledge that 2013-03 conflicts with ARIN's need requirement for inter-region transfers, and that this is a valid argument against the proposed policy, I do feel that this argument alone is not strong enough to to stop 2013-03, for the following reasons:
1) It is entirely dependent on proposal 2012-02 passing, which is far from a certainty. 2012-02 has received at best a lukewarm response from the community - according to the chairs, it has, quote, «far from consensus». Should 2012-02 not pass, 2013-03's conflict with ARIN's need requirement is completely irrelevant.
2) According to Ingrid Wijte's MENOG12 presentation ( http://www.menog.org/presentations/menog-12/127-IPv4_Transfers-RIPE_NCC_Upda... ), there have been only a measly 17 permanent transfers in the RIPE region in the last three months. It would surprise me greatly if the amount of assignments having been made by the LIRs in the same period is not at least two orders of magnitude more.
I am somewhat baffled by this. Do you (or does anyone) have a good idea as to what ISPs are doing instead? Are they improving utilization efficiency within their network? Still using up space they got before the free pool was exhausted? Or (I doubt this) putting new users on IPv6? I know that the RIPE intra-RIR transfer policy is somewhat strict compared to those in APNIC and ARIN (not allowing transfers to end users, for example): is that likely playing a part as well?
Choosing to uphold the need justification and documentation bureaucracy for all those assignments that we LIRs perform on a frequent basis, for the sole purpose of allowing only one specific flavour of the very rarely used allocation transfer mechanism to work, seems to me to fail a very basic cost-benefit analysis.
3) The RIPE community's benefit of allowing transfers from the ARIN region into the RIPE region is only obvious due to the fact that ARIN still has remaining free space in their address pool. «They've still got more space, let's go grab as much of it as we can before it's all gone!»
It is not certain that it is a net benefit to the RIPE community to allow transfers with depleted regions. The result may very well be that more address space is being transferred *out* of our region, than what is coming in - further exacerbating our depletion ordeal. This uncertainty makes the cost-benefit analysis I mentioned in #2 even more clear in favour of 2013-03.
According to Geoff Huston, ARIN looks set to deplete exactly one year from now. Keeping in mind that 2012-02 appears to still have a long way to go in the PDP, the time period during which the benefit of allowing free trade with ARIN is obvious is only going to be a few months at most. Quite possibly, ARIN will deplete before 2012-02 gets implemented, if so the obvious beneficial period will not exist at all.
The inter-RIR transfer policies were specifically designed to prevent the free pool of one region from being drained via transfers to another region. Rather, the benefit of inter-RIR transfers comes from getting existing underutilized space back into use. As it happens, the ARIN region has a lot of large legacy allocations, many of which are underutilized. That source of supply, IMO, is what the RIPE region will be getting access to if/when a compatible inter-RIR transfer policy is passed. It is worth noting that prices in the transfer market to date are significantly higher in the RIPE region than the ARIN region. I believe this reflects the relative scarcity of supply in the RIPE region (which is likely also related to the reason there have been so few transactions). In the ARIN region, I believe prices already reflect sellers' expectations of the post-exhaustion supply/demand balance (non-bankrupt sellers are mostly holding out for free pool exhaustion rather than selling at a lower price). As a result, I don't expect transfer prices in the ARIN region to rise significantly when that demand materializes.
4) Again, ARIN is soon depleting. As I mentioned in the proposal itself, this might cause the ARIN community to do a reality adjustment similar to 2013-03, and rescind the need requirement for inter-region transfers. After all, there is little point in trying to prevent other regions from "bleeding them dry" if they have nothing left anyway. So that's your item #3, essentially. If you're willing to suggest that to the ARIN community now, I'm all for it.
We will undoubtedly discuss this (at least informally with interested parties) at next month's ARIN meeting. I suspect that if the discussion gets any traction, any resulting proposal will be to relax needs assessment requirements on transfers and reassignments only after the ARIN free pool is exhausted. -Scott