[ppml] Policy Proposal: Decreasing Exponential Rationing of IPv4 IP Addresses

[Iljitsch van Beijnum]

On 27-aug-2007, at 23:06, Dean Anderson wrote:

> Fact 1: The Available IP Pool (AIP) will be exhausted on or about  
> March,
> 2010 if no changes are made.

Geoff Huston predicts june 2010 for the IANA pool and march 2011 for  
the RIR reserves. This requires a rather significant increase in  
yearly address use. (2005 and 2006 were about 170 million addresses/ 
year, we are now at around 195 million for the last 12 months,  
running out in 3.6 years means 325 million/year average during that  
period.) In my opinion, the data is too volatile to support such a  
prediction. Something like 2012 - 2013 seems more reasonable. Then  
again, I'm not a mathematician.

> Fact 2: If the AIP is exhausted, it is unknown when address space will
> be returned that can be delegated.

Address space is returned all the time. Last time I check was two  
years ago IIRC, and that was 11 million addresses in a year. I'd say  
it's likely that this will go down as we run out of IPv4 space  
because people will want to hang on to what they've got and returning  
a few small blocks for a larger one won't happen anymore.

It goes without saying that the returned address space won't be  
nearly enough to cover the requests that come in.

> Fact 3: Disruptions of unknown duration are more harmful to
> business planning than disruptions of known duration.

Even more harmful is thinking a disruption is temporary when it is in  
fact permanent.

> Fact 5: Rationing any kind of limited resource inhibits hoarding

No, it just makes it start sooner and the iterations as new blocks  
come available and assignments resume temporarily allow the hoarders  
to learn and become more effective.

> Fact 6: IPv4 usage will not end because of AIP exhaustion.

Growth will end, and many economic models require growth.

> Fact 7: IPv4 allocation will eventually resume after exhaustion

Not in any meaningful way. Sure, if you need a /24, you'll be able to  
get it at some point, but if you need a /12, that's simply not going  
to happen.

> However, after exhaustion, we won't be able to
> predict when allocation might resume.

It's not a stop/wait/resume thing, what will happen is that large  
blocks will no longer be available but small blocks will continue to  
be assigned unless so many requests come in that seem legitimate  
enough to soak up all remaining space.

> Your reason for objection is to promote IPv6.  IPv6 should have  
> nothing
> to do with IPv4 resource management.

A smooth transition from IPv4 to IPv6 is in everyone's interest. IPv4  
is a sinking ship. You don't have to jump in a life boat right this  
minute, but planning new trips is not a smart thing to do.

>> Yes, and after all the cars have been turned into big piles of rust,
>> you'll be able to get oil again, too. But that doesn't mean things
>> revert back to the way they were before.

> We won't have to wait thousands of years for IPv4 address to become
> available again.

Cars rust faster than that.

I don't mean wait for new oil to be formed again (that would be a  
while) but there will always be SOME oil and without cars to burn it  
up, there will probably be enough for many other uses, such as making  
plastics.

> There is no expiration date for IPv4.  That's your fallacy (or  
> fantasy).

Right, when 9 billion people are running IPv6, the 900 million people  
running IPv4 have no reason to change protocols.

>> I'm not trying to wreck IPv4.

> I think you are. You are attempting to stop prudent action to avert a
> foreseeable and avoidable problem with a shortage in IPv4 IP Address
> Availability.  That doesn't help keep it in a "usable state".

Your mistake is focussing on IPv4 exclusively. The point is that we  
have a working internet. Today, we need IPv4 for that. In a few  
years, IPv4 won't be able to provide this function for new users, so  
we need to move to IPv6. Making it harder for new users sooner  
doesn't help those IPv4 and the market forces are such that it also  
delays the real solution = moving to IPv6.

> Your justification for your view is that you want to get IPv4 'over  
> and
> done with' to move onto IPv6.  That's just sabotage of IPv4.

Absolutely not. I want IPv4 to remain useful as long as it can, which  
means NOT making it harder to get IPv4 address space, which is  
already relatively hard today.

>> China has gotten 30 million addresses so far this year. That's more
>> than all year last year. Once China has caught up, the really poor
>> countries are next in line. IPv4 can't sustain our growing
>> communication needs, it's as simple as that.

> Then they'll be happy to move to IPv6. You won't have to mismanage  
> IPv4
> to encourage them.

Oh so it's only the poor people that should use IPv6?

>> I recommend reading a good book about it.

> Its kind of hard to find a good book that hasn't been obsolesced by  
> the
> continuing changes to IPv6, judging by the RFC index.

Anything published in 2003 or later should be in reasonable shape in  
that department. My book is now 2 years old and I don't think there  
are significant issues in this area. Sometimes it helps that the IETF  
moves so slow.  :-)

> Here's one DNS example: Recently, a Country code (cctld) domain  
> operator
> tried to add IPv6 AAAA record support to their cctld name service.  
> If I
> recall correctly, they found that they could only have 6 A records,  
> and
> 3 AAAA records before they ran out of the (IPv4) packet size  
> limitations
> (512 bytes).

EDNS0 is your friend, people who use stupid firewalls get what they  
deserve. I don' see a problem. (But then again, I'm not a root server  
operator.)

> There is no technical reason that IPv6 DNS should talk to an IPv4
> nameserver.

The DNS is used to determine if a remote system is reachable over  
IPv4 or IPv6. This means that address records for both protocols must  
be present in the same name space and then separating the two  
protocols makes no sense.

> So, there is a lot of cruft and limitation in IPv4 DNS that
> would be quite good to remove from IPv6.  None of that happened.

I'm with you on that one, they could have made EDNS0 mandatory for  
IPv6. Although I think the real issue here is the IPv6 socket API,  
which happily continues the layer violations that were unavoidable  
for IPv4 but could have been cleaned up when IPv6 required changes.

> IPv6 is designed in someways to profit some people.

Yeah right. People are making money off of IPv6 left and right...

> The other part is that IPv6 seems to be so unstable (due to continuous
> changes) that nothing works very well.

Actually it works extremely well. There are lots of times when IPv4  
requires hacks that are completely unnecessary with IPv6.

Unfortunately, IPv6 doesn't (yet) do everything IPv4 does. For  
instance, you can't do dial-up over IPv6 in a mixed vendor environment.