As we run out of IPv4 addresses, should the Regional Internet Registries allow address blocks to be traded? In a paper released last month, we argued that a liberalized transfer policy would serve a number of important purposes. Transfer policy proposals are quite controversial, however. In an attempt to gauge levels of support, the American Registry for Internet Numbers (ARIN) has taken a survey of its public policy mailing list. About 200 people responded.

Overall, the survey found that 87% of the respondents supported a liberalized transfer policy of some kind. It is a level of support much larger than one would have anticipated from the debates on the lists, which seem to have been dominated by vocal opponents. Another result showing support for a more liberal policy pertains to controls on speculation. In ARIN, RIPE and APNIC, the draft transfer policies all imposed on the buyers of addresses two year time limits on re-selling the address resources they received. These time limits were put in place to control speculative trafficking in address resources. The ARIN survey asked “how long must a recipient of space under the policy hold it before space can be transferred to another party?” Only 28% supported a two-year limit; an impressive 24% supported no limits at all, while 37% supported a limit of only one year. It would be safe to say that the two-year limit does not have much support.

The more liberal policy toward speculation might be a result of the strong support for continued “needs assessment” by ARIN in advance of any sales. It is possible that restricting buyers of addresses to those who claim to “need” them would limit speculative accumulation of address blocks – although loopholes could almost certainly be found. By a ratio of 72% to 28%, a majority believes that ARIN should “pre-qualify” organizations before they can buy addresses. However, the success of such an approach will really depend on how ARIN defines “need.” Many in the RIRs naively believe that need is an absolute and objective parameter that can be ascertained via engineering studies. In reality, need is a largely economic matter contingent upon price, supply, demand and alternative technologies. Let’s hope ARIN opts for a more flexible approach to needs assessment.

More disturbingly, there is a 50-50 split on the question of whether organizations trying to sell addresses to others should be “pre-qualified.” We confess that we are unsure what it means to “pre-qualify” a seller. It seems self-evident that in an environment of increasing IPv4 address scarcity, any organization wants to give up their addresses to a willing buyer should be encouraged to do so. If “pre-qualification” simply means that sellers must show that they are actually the legitimate holder of address resources they are selling, then it is a good idea, but it would be better to call it “authentication.” If it means that sellers must show that they somehow had a need for the addresses they are trying to sell, then pre-qualification is utterly stupid and self-defeating. It suggests that ARIN will expropriate or somehow punish organizations that have too much address space and are trying to transfer it to others. All such a policy will accomplish is to prevent many address block holders from ever coming to the table, which keeps the resources lying fallow and defeats the whole purpose of a transfer policy.

It will be interesting to see how ARIN handles the split around this question. ARIN, like other Internet technical organizations, drinks deeply of the Kool-Aid of “consensus policy making” by some mythical collective entity called “the community.” Either ARIN pre-qualifies sellers or it doesn’t – ain’t no middle ground there. When 48.5% of “the” community says “Yes, do it;” 46.5% says “No, don’t do it;” and 5% says “I Don’t Know,” what happens? We'd suggest that sellers be authenticated, not “pre-qualified.” If there's no preponderance of support for a policy there should be no policy.