[arin-ppml] IANA IPv4 /8 burn rate.... (was Re: Steppingforward, opening my mouth and removing all doubt about)

[Iljitsch van Beijnum]

On 28 aug 2008, at 20:22, Milton L Mueller wrote:

>>> of course everyone will use a lot of them.  When they become scarce
>>> and expensive, people will start conserving IPv4.

>> How is that a good thing?

> When you don't know how long it will take to convert, and are  
> uncertain
> about the cost of migration, stricter conservation allowing for a  
> longer
> transition, if needed, is prudent policy. If a longer transition  
> period
> is not needed, a transfer policy does no harm, as the resources become
> devalued more rapidly.

You assume that having IPv4 space available is a good thing, which I  
can agree with, but also that stricter conservation isn't harmful, or  
the harmful effects are less than the good that comes from having IPv4  
addresses available.

I disagree with that. The current rules are already so strict that  
they're harmful. The inability to get addresses is a huge problem. See  
the NAT issue. I agree that NAT works for most people most of the  
time, and we can argue about how often it doesn't work for how many  
people, but it's obvious that it breaks stuff for some people. So  
these people would be better off not using NAT, which means they need  
more IPv4 addresses. They usually can't get them at a price that they  
find affordable. Today, that's mostly due to ISP policies. Giving out  
more than a single IPv4 address is significant extra work and you can  
probably make more money by keeping the prices for this extra service  
high than by making it cheap.

Also, an address is only of use (and value) when someone uses it.

>> IPv6 addresses are free, because whatever the cost is, after dividing

> Business/economic reality check: nothing is free. Certainly not a
> protocol that requires heavy investments in new equipment and human
> resources/skills.

Sure, the switch to IPv6 isn't free. (The addresses are.) But you need  
to train people and upgrade equipment periodically anyway. So if you  
take long enough, the switch to IPv6 doesn't cost all that much money.  
However, if people prefer to stick their heads in the sand and then  
make the switch over night, that's going to be very expensive.

>> What we should do is try to make the switch to IPv6 as painless as we
>> can. The most important part of that is to make it predictable: we
>> need to know what's going to happen in the next 5 years or so.

> Interesting. You think we can make the migration "predictable" by
> erecting a brick wall.

The brick wall is already there: 3706.65 million usable addresses,  
992.46 are still available. We get closer to the brick wall at a rate  
of about half a million addresses a day.

> One logical consequence of this argument: no need
> to wait, we can do it tomorrow, or as soon as ARIN agrees to adopt a
> flag date where IPv4 just ends. Agree?

1. No: people wouldn't do it just because ARIN (or anyone else)  
mandates it
2. Obviously there are other costs than just unpredictability

> Can you _really_ predict, with reasonable levels of probability, that
> you will even prevent private unauthorized transfers from taking place
> under those circumstances?

Of course not. IP transit service usually comes with IP addresses.  
Unless the customer has a portable address range of their own, this is  
a necessary part of the service. Now this wouldn't be "buying" but  
more like "leasing" but that doesn't really matter in the grand scheme  
of things.

The point is that IP addresses are a limited resource and everybody  
should have the same access to that resource. Some people getting it  
for (almost) free from ARIN and others paying a lot, but without  
conforming to the ARIN rules would be a bad outcome, because the ARIN  
rules are such that the level of address constraining pain is kept  
equal across the industry.

Some rules only work if they are applied 100% all the time, but in  
most cases they are still beneficial if a few people break them, as  
long as this doesn't get out of hand. So the fact that a few people  
manage to get addresses in ways that they shouldn't doesn't  
automatically mean that we should remove the rules.

Of course we have a small number of organizations that manage to hold  
a lot of space without conforming to the rules (the legacy holders,  
with the most prominent among them the US government with some 10 /8s)  
and ARIN and/or its counsel have conceded defeat there, which was a  
very bad thing. Fixing that by bribing these people to give up their  
address space won't make it right.

> But why not go whole hog with the uncertainty argument: let's just  
> get a
> decree to stop operating the IPv4 internet tomorrow. That would really
> clear things up, eh?

There is an important difference between not being able to deploy new  
address space and breaking the already deployed IPv4 installations.

>> Any policy change means that there will be a bigger difference  
>> between
>> what's being predicted and what will happen,

> I'm afraid that that's not true. You are assuming that the migration  
> is
> 1) a linear, progressive and predictable process and 2) will occupy a
> known time period. Both assumptions are shaky.

Nobody knows what will happen as we run out of IPv4 space. But as long  
as we can predict that this will happen with enough certainty that  
people can't ignore it (which is a point we're finally reaching right  
about now) everyone will be forced to start planning for a future  
where no new IPv4 space can be obtained. Presumably, some people will  
botch this and have a problem, but most people will come up with a  
strategy that works for them so the internet will continue to work  
after we run out of IPv4 space.

> A transfer policy is a hedge. It maximizes efficient utilization of
> remaining v4 resources while in no way precluding IPv6 migration.

There was nothing precluding IPv6 migration five years ago. But nobody  
migrated. As long as people keep squeezing the IPv4 toothpaste tube  
they won't get around to buying a new tube and their teeth will rot in  
the meantime.

So: let's use IPv4 the way we've been doing while it lasts and plan  
for a no-new-v4 future in the mean time.

> In an environment of IPv4 scarcity, with large swaths of v4 address
> blocks known to be underutilized, and with occupiers of address
> resources needing an incentive to migrate, allowing holders of those
> address blocks to financially benefit from releasing them is an  
> obvious
> win. That's the only reason I support it.

No. I'm 95% sure a functioning address market for IPv4 addresses won't  
form for the 1M+/yr address users (= all the broadband ISPs) because  
for them even a few dollars per address is more expensive than the  
alternatives while freeing up large address blocks is expensive.  
(You'd probably have to audit all of HP's IT infrastructure to see if  
there is old DEC gear that still has a 16.x.x.x address before you can  
sell off 16/8, for instance.) So we'll have a lot of confusion,  
experiments and waiting going on and two years from now we're still  
pretty much in the same boat except that then we have to make the  
transition in half the time that we have today.