Some observations on the differences in the various transfer policy proposals
Hi, At the ARIN meeting last week the question arose as to why the various policy proposals related to address transfers in the different RIRs were so different. I made some comments in response to this question from my perspective and then I had some followup questions mailed to me, so I thought maybe there is some value in writing my perspective up. If you are interested its at: http://www.potaroo.net/ispcol/2008-11/transfers.html thanks, Geoff Huston DIsclaimer: I'm speaking for myself, again!
Geoff, I've read your article and thanks for sharing your observation. Your classification of registry's role and how it will change before and after the IPv4 exhaustion(RIRs' pool becomes unavailable) was interesting and helpful in clearly identifying what the remaining role for RIR will be. While I agree that effectiveness of regulation is linked to specific benefits (in the case of current situation, ISPs follow RIR policies because they need to receive subsequent allocations from RIR), I'm not sure how RIRs' role of regulation will completely disappear after IPv4 exhaustion. If transfer will be the new mean to obtain additional IPv4 address space, and if RIR is the authority to register transfer records, wouldn't this act as an incentive for LIRs to follow RIR policies? RIRs' role simply switches from making allocations to reconizing transfer, i.e. LIRs will not be allowed to transfer their records (obtain IPv4 address space) if they don't follow RIR policies. You mention that there may be alternative registries that provides easier transfer without any contraints, and some people may indeed set up such registry, but I don't quite see how people will immediately trust an alternative registry to replace RIR registry. To put it in short, as long as some means of obtaining IPv4 address is accomodated (in this case, transfer of already distributed space), I'm not sure if people will go as far as breaking as in your assumption, even if some degree of contraints are added. Just my thoughts on reading your paper. izumi JPNIC Geoff Huston wrote:
Hi,
At the ARIN meeting last week the question arose as to why the various policy proposals related to address transfers in the different RIRs were so different. I made some comments in response to this question from my perspective and then I had some followup questions mailed to me, so I thought maybe there is some value in writing my perspective up.
If you are interested its at: http://www.potaroo.net/ispcol/2008-11/transfers.html
thanks,
Geoff Huston
DIsclaimer: I'm speaking for myself, again!
Hi Izumi, Thank you for your comments. As always, making reasonably accurate predictions of the future is hard, and setting up policies that will be effective in the future is very challenging. One possible way is to look at "what if" scenarios and see how particular frameworks could survive various potential events and see what the implications may be. What if... the registry imposes the same constraints of demonstrated efficiency of utilization of existing addresses and demonstrated need on access to the registry, and alternate registries do appear. What then? Can the registry pull these alternate registries back into a coherent whole again? How could that possibly happen? Where is the authority that says "this registry is the authoritative one! You MUST ignore all others?" I personally simply cannot see how such a situation is resolvable in any useful manner. And what happens to a market function in the face of such attempts to impose constraints on buyers and presumably on sellers? As far as I can see this path poses excessive levels of risk to the integrity of the registry function, the coherence of addresses and to the survival of the Internet itself. The problem is that we will continue to need a working fully functional IPv4 network throughout this transition ot IPv6 and if we break that IPv4 network its hard to see that "just use IPv6" is a useful response. What if... the registry simply registers transfers without imposing extraneous constraints on the registrants. i.e. the registry simply says "yes"? What is the downside here? The most commonly cites risks that I have heard is routing table bloat and the formation of a market in addresses. In the case of the routing table, the observation I would like to make is that there has really never been any constrain on the routing table so far and the routing table continues to be populated in a manner that 50% of the routing table are more specifics and this has been the case for years (http://bgp.potaroo.net/cgi-bin/plota?file=%2fvar%2fdata%2fbgp%2fas2.0%2fbgp-spec-perc.txt&descr=Specifics%20as%20%25%20of%20FIB%20entries%20(153312%2f285680)&ylabel=Specifics%20as%20%25%20of%20FIB%20entries%20(153312%2f285680)&with=step ). So my response would be that any notion that registry policies and constraints would have any effect whatsoever on the routing table size would have to address the apparent contradiction that so far the policies related to allocation have had effectively no effect on the routing table bloat to date. And what about markets? Markets will form in any case. And, as I understand conventional thinking relating to market behaviour, the pricing function is a form of equiliberation between buyers and sellers. And open market allows potential sellers to have exposure in front of the largest possible pool of buyers (a positive aspect for market behaviour I thought) and buyers to be able to see the entire set of sellers (again a positive aspect). What is the price escalates? Well as has been pointed out before such a movement acts as a clear market signal about the imperative to transition to Ipv6 _while the Ipv4 network is still a coherent and functional network to support the transition_) So I see the notion that "we should change as little as possible and we should cling desperately to our cherished allocation policies even when there is nothing left to allocate" as not being a conservative notion, nor even a quaint and amusing notion, but an astonishingly radical and extremely risky notion. In my view attempting to tie the by-then anachronistic allocation policies to the registry function to be a course of action that imperils the coherence of the address system for the entire Internet, imperils any hope of transitioning to IPv6 and provides impetus to the industry at large to spin off into all directions to try and save their own investments and own customer base. At that point there is no Internet left. I just cannot see why such a radical and risky course of action can be seen as being "conservative" or "safe" when to me its entirely the opposite! Geoff Disclaimer: These are, as always, entirely my own views on this situation.
participants (2)
-
Geoff Huston
-
Izumi Okutani