[ppml] IPv6 assignment - proposal for change to nrpm

[Stephen Sprunk]

Thus spake "Steve Bertrand" <steveb at eagle.ca>
>> Right.  One of the goals in IPv6 policy is minimizing routes in the
>> DFZ, and it's easiest to do that if everyone (or nearly everyone)
>> has the same size blocks because it's easy to filter deaggregates
>> that way.  While a /32 is way, way too large for most LIRs, it's
>> large enough that nearly all LIRs will fit in it, it's a convenient
>> number, and everyone can filter anything longer than /32 (except
>> in the PIv6 block, where it's /48) with impunity unless they
>> specifically want deaggregates.  Since we have absolutely no
>> clue how to route the half a billion /32s in 2000::/3, there is no
>> reason to give out longer prefixes -- and plenty of reasons not to.
>
> I completely understand the importance of the allocation of blocks
> with as short a prefix and as consistent as possible, but thanks for
> the clarification.

"As short as possible" wasn't a consideration, AFAIK.  We certainly don't 
want any risk of running out of them, which is what happens when you get 
stingy with the size of a protocol field.

> Out of curiosity, who was the original classification of size (/32
> and /48) distribution designed by?

That came from the IETF, though it's fairly obvious where they got it from: 
multiples of 16 bits are much preferred by humans given how v6 addresses are 
written, and powers of two (and small multiples thereof) are preferred by 
computers given register sizes -- the same sorts of reasons that led to 
8-bit boundaries in v4.  Once EUI-64 was decided upon, I can't see how we'd 
have divided 128 total bits any other way than was done.  If they'd gone 
with EUI-48, it'd probably look slightly different, but not too much.

> Was it the community as all other ARIN policies are created by?

One can debate how much the IETF resembles "the community", but the RIRs 
have a long history of following the IETF's recommendations, with a few 
notable exceptions.  Also, a number of folks from the RIR world are involved 
in the IETF world as well, which keeps the two reasonably in sync most of 
the time.

> Also, I don't pay attention to what other RIR's are doing, is this
> allocation scheme the same across all RIR's?

Initially, all the RIRs started out with a uniform policy set. 
Subsequently, each has monkeyed with policies in different ways so they're 
no longer identical -- just as happened with IPv4.  ARIN's major v6 policy 
changes to date are the addition of PI /48s and PA /56s and the removal of 
non-policy text (the latter mainly due to our different policy format).

S

Stephen Sprunk         "God does not play dice."  --Albert Einstein
CCIE #3723         "God is an inveterate gambler, and He throws the
K5SSS        dice at every possible opportunity." --Stephen Hawking