[arin-ppml] Set aside round deux

[michael.dillon at bt.com]

> Let me offer a rude viewpoint to gauge reaction: the 4.10 addresses
> shouldn't be available to the X-larges at all. Period. The X-larges
> have vast tracts of IPv4 addresses from which they can find a few to
> facilitate IPv4 function during the v6 transition. 4.10 addresses
> should be for folks who didn't have a lot of v4 addresses to start
> with and need just a few more to carry them through to v6 ubiiquity.

What's rude about it? This is reality. All large ISPs have enough
existing IPv4 space to keep their businesses humming even if they
can't get more from ARIN.

That doesn't mean that they currently have spare addresses stashed
away, although I would expect that most do have sloppy decommissioning
processes with the end result that they have addresses lost in the
system. But the main source of IP addresses for these large ISPs is
their existing customers. They will have high margin customers and
low margin customers. It is not unusual for a large company to identify
its low margin customers, and then contact them all to say that 
their service contracts will not be renewed and if the customer
wants to cancel the contracts at any time, no penalty will be 
charged. 

Therefore, the large ISPs are able to mine their low margin customer
base to recover IP addresses which can be used for continued growth
of high margin customers. In some cases this kind of process will have
other benefits. For instance, consider an ISP who installed lots of
75xx edge routers to sell T1 access. Over the years, the high margin
customers have all upgraded. By shedding certain low margin customers
they not only recover IP addresses, but they can get rid of those
75xx boxes freeing up rack space for 10Ks with a higher port density
and greater flexibility. 

I believe that address recovery through internal churn is viable for 
all of the larger ISPs therefore I think that a policy which intends
to limit the disruptive effect of IPv4 exhaustion would do better to
focus exclusively on the smaller businesses who often have less 
flexible business models. I'm not saying that ARIN has a responsibility
to protect these businesses from bankruptcy or buyout, but that a
policy focused on lengthening the period of time during which these
businesses can get IPv4 addresses, will generally ease the process
of IPv6 transition. For one thing, it will make it more likely that
these smaller businesses can buy IPv6 routers on the used market.
And it will allow them to spread their investment in transition
over a longer period which makes it more financially feasible.

The basic principle of set-aside policies is that ARIN will impose
pain on certain organizations in order to protect others. The goal
is to make the IPv6 transition a bit smoother. Which orgs can handle
the pain? The big ones. Which orgs would get most of the pain anyway
if ARIN does nothing. The big ones. The Robin Hood strategy works
for me.

--Michael Dillon