[ppml] IPv6 assignment - proposal for change to nrpm

[Scott Leibrand]

John Osmon wrote:
> Two questions hit me while I was reading this thread:
>  -- hoarding -- is it hoarding to keep the address block you
>     were assigned, even it if is bigger than you wanted?

I wouldn't think so, at least in the case of someone who got a /24.  If 
you got a /8 or /16 and are only using part of it, I think the proper 
thing to do is to renumber (if necessary) into a subnet of your original 
allocation, and return the unused portion.

>   How many legacy holders would have been happy with a /26 - /29 if
>     such blocks had been available?
>   

Given the way routing filters played out (at /24), I would say not many.

>  -- address run-out -- How much longer could the existing address
>     space last if we could allocate a block of space that is
>     appropriately sized for an end-user's needs?
>   

I think this can, will, and is happening.  All new allocations for the 
last 10 years have been appropriately sized.  Some who received /8's 
have already renumbered and returned un-needed space.  I'm hopeful that 
this will continue happening with other legacy /8 and /16 allocations 
and assignments.  I also suspect that as we approach IANA free pool 
exhaustion we'll be able to implement an address transfer system, 
whereby networks needing new space can incent networks with existing 
space to renumber into smaller netblocks.

> Yes -- I realize that I'm playing fast and loose with other people's
> routing slots.  Which is worse?  Running out of address space?  Or
> running into limitations of current router implementations?
>
> If we play towards the router implementations, aren't we setting
> routing policy de facto?

Clearly both can be bad, and any policy we implement needs to 
appropriately balance the two.  That is why I favor some sort of address 
transfer system, but also think we need to maintain restrictions on the 
top-level deaggregation of assignments and allocations.  I believe that 
deaggregation at lower levels of the hierarchy is less of a problem, as 
that preserves the possibility of filtering deaggregates while 
preserving reachability via the larger covering aggregate.

-Scott