Re: [ppml] Policy Proposal: IPv6 Assignment Size Reduction

["Stephen Sprunk" <stephen@sprunk.org>]

Thus spake "David Conrad" <david.conrad@icann.org>
> On 10/30/07 2:53 PM, "Brian Dickson" <briand@ca.afilias.info> wrote:
> > I have, however, argued that ISPs returning to RIRs for PA blocks, is
> > bad for the Internet.
>
> Only if those new PA blocks can't be aggregated.

... and only if there's a non-trivial number of them.

Realistically, if LIRs are handing out /48s (or /56s, grr) to every 
customer, there's only a handful in each region that would run out of space 
in a /32 and would need to come back for more -- which was the point, IIRC. 
Plus, those are the exact same folks that are asking for (and getting) more 
than a /32 to start with.

> > Avoiding repeat PA assignments to ISPs, is the *primary*
> > motivation for the proposals, both at the IETF (6man) and ARIN
> > (ppml).
>
> The simplest way to do reduce the impact of multiple PA assignments
> would be for the RIRs to assign address space in a sparse fashion
> such that additional allocations could be done by simply subtracting
> a bit from the allocated prefix length.  In fact, the RIRs promised to
> do this when the /12s were allocated by IANA (last I checked,
> however, only APNIC was doing it).

Yes, allocating via bisection would help with that quite a bit.  Currently, 
it appears ARIN is allocating /32s out of a reserved /29 for each LIR, which 
allows 3 bits of growth for each before a second prefix hits the DFZ.  That 
doesn't seem to be enough at first glance.

However, the few mega-ISPs out there are asking for blocks larger than a /32 
(e.g. /27 and /29), and those larger blocks appear to be coming out of 
reserved /21s (in ARIN's region).  That's a lot of bits for growth for the 
folks that indicate they don't expect a /32 to be sufficient in the long 
run.  If that still turns out to be insufficient, I'm not so worried about 
AT&T, VZB, et al announcing two or three aggregates as I am them announcing 
hundreds to thousands of intentional (e.g. TE) deaggregates.

Are we trying to solve the right problem here?

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