[arin-ppml] Draft Policy 2011-1 - Inter-RIR Transfers - Shepherd's Inquiry

[Joe St Sauver]

Regarding "Draft Policy 2011-1: Inter-RIR Transfers"

and the proviso that:

   "Address resources may be transferred.... in or out of the ARIN region
   to those who demonstrate need and plan to deploy them for a networking
   purpose within 3 months. Such transfers will take place between RIRs who
   share compatible, needs-based policies on behalf of entities agreeing to
   the transfer and which otherwise meet both RIR's policies. Transferred
   resources will become part of the resource holdings of the recipient RIR
   unless otherwise agreed by both RIRs."

I am opposed to inter-RIR transfer policies, including this one. Let me
mention just a few reasons why.


I. Resource Firebreak Issues.

In thinking about this and similar policies, consider where addresses would 
likely come from and where addresses would probably go given current 
circumstances (and assuming that all RIRs agreed to participate in the 
policy).

RIRs with remaining IPv4 address resources ("supplier RIRs")

-- AFRINIC
-- LACNIC

RIRs with a likely immediate-term demand for additional IPv4 addresses
("consumer RIRs"):

-- APNIC
-- RIPE NCC
-- ARIN

Thus the flow of address space would thus likely be *from* the developing
nations of the southern hemisphere (except Australasia) *to* the developed 
nations of the northern hemisphere.

Countries in the southern hemisphere (which have historically been handicapped 
by the high price of connectivity) have just recently begun to see improved
fiber access and a resulting drop in market prices. With their large areas 
and large populations, they will likely have a substantial demand for number 
resources in most forseeable medium-term scenarios. If the critical resources
they will eventually require will have been "exported" piece meal to Asia, 
Europe, or North America, our southern-hemisphere neighbors will find it hard 
to realize their full potential as Internet citizens.

The existing region-based RIR system thus serves as a "firebreak" of sorts, 
protecting each RIR's resources from being treated as one common comingled 
pool that can be (effectively) drawn down in its entirity by any region
willing to do so.

I don't want to see that "firebreak," small though it may be, breached. I 
think it serves an important purpose when it comes to protecting resources 
in developing regions such as AFRINIC and LACNIC.


II. Additional Resources Should Not Be Allowed to Potential Feed Regions 
with Unchecked Abuse Issues.

Some have also argued that RIRs vary substantially when it comes to 
their interest in policing number resource misue and abuse. An interesting
exercise when it comes to measuring one dimension of that issue is to go to 
Spamhaus and see how many SBL listings are associated with each of
the RIRs. At the time I checked just now:

-- http://www.spamhaus.org/sbl/listings.lasso?isp=afrinic
   4 listings

-- http://www.spamhaus.org/sbl/listings.lasso?isp=apnic
   18 listings

-- http://www.spamhaus.org/sbl/listings.lasso?isp=lacnic
   28 listings

-- http://www.spamhaus.org/sbl/listings.lasso?isp=arin
   250 listings

-- http://www.spamhaus.org/sbl/listings.lasso?isp=ripe
   I quote, "has far too many records to list. This ISP has an extremely
   serious spam problem."

Although I've looked at a LOT of SBL listings over the years, this is
absolutely the first and ONLY time I've EVER seen a "too many to list"
response from the Spamhaus by-ISP SBL listings.

Thus, I would also oppose inter-RIR transfers if those transfers might
potentially serve to enable further messaging abuse by supplying "raw 
materials" to areas where number resources are apparently experiencing 
rampant abuse. (If all or many of those issues have been taken care of, 
and Spamhaus is just running behind on removals, my apologies)


III. Tracking and Documenting Address Usage Becomes Harder.

Currently it is possible to track address usage because we know (at 
a /8 granularity) which regions have a given block. If we suddenly
begin allowing /16s here and /19s there to be carved off and transfer
out or transfered in, our ability to track usage by region will quickly
be lost (or become a tiresome and highly granular bookkeeping exercise).

Moreover, IP whois currently relies on that assignment by /8s to 
direct IP whois queries to an appropriate RIR whois server. If inter
region transfers are allowed, whois queries will now potentially need
to support still more redirection, and protection against whois 
redirection loops will likely even become necessary. And do we want 
the RIRs to need to update their whois to deal with (redirect) thousands 
or tens of thousands of tiny blocks that may be transfered out of region? 
Jeez, again, what a bookkeeping mess!


For all these and other reasons, I oppose inter-RIR transfers, including
the current proposed policy.

Thanks for considering these comments,

Regards,

Joe

Disclaimer: all opinions expressed are strictly my own and do not 
necessarily represent the opinion of any other organization or entity.