Re: [sig-policy] IPv4 countdown policy proposal

[Takashi Arano <arano@inetcore.com>]

Hi David,

At 15:24 07/02/27, David Conrad wrote:
>Arano-san,
>
>On Feb 26, 2007, at 1:43 AM, Takashi Arano wrote:
>>>- Currently, APNIC has an "80% rule".  Create a policy that when the
>>>current free pool is reduced by 50%, make it a 90% rule.  When the
>>>remaining free pool is reduced by another 50%, make it a 95%  
>>>rule.  Etc.
>>I don't believe it doesn't lead to significant conservation.
>
>The point of this example was to make it self-adjusting.  As the free  
>pool gets smaller, the restrictions increase automatically.  By  
>definition, it would extend the free pool as far as it needs to go --  
>the restrictions would get arbitrarily complex so at some point it  
>becomes easier/cheaper/more cost effective to simply deploy IPv6 and  
>the NAT boxes necessary for IPv6-only sites to talk to the rest of  
>the Internet.

Of course, I believe I understand this.

I don't just understand your example does not seem effective very much
for conservation.

Assume that LIR A is consuming IPv4 addresses constantly and 
they are given /8 of allocation for one year's usage. Here 80% rule
results in sebsequent request after ten months when they consume
80% of /8. In 95% rule, they consume 95% of /8 in nearly 12 months
and then come back to registries for subsequent addresses.
In both rules, LIR is allocated /8 per a year. 95% rule just delays
the timeing of subsequent request a little bit. 
This is why I think your example is not so effective.

I would agree that mix and match may be useful. But even if
we succeed in conserving 20%, exhaustion date would be
prolonged just one year, where LIRs would have severe time
being allocated. If you like gradual policy, probably that policy 
should conserve more addresses, say 40% which can prolong
address life time from 5 years to 8 years or so.
In order to do so, I guess policy should be more drastic
and we would have to discuss who and in which cases
addresses are not allocated/assigned.

Doesn't it make sense?

Regards,
Takashi Arano

>>>I'm sure there are many others.
>>Again, I don't think there are many.
>
>There are a myriad variations and combinations that can be applied.
>
>>It would be very difficult to decide which portion of address blocks
>>that are currently allocated/assigned will not be allocated/assigned
>>in a new policy.
>
>I must not understand this comment.  I would assume the new policy  
>would apply to the free pool existent at the time the policy was  
>implemented, just like policies have always been applied.
>
>Rgds,
>-drc