[arin-ppml] Encouraging IPv6 Transition

[William Herrin]

On 5/16/12, John Curran <jcurran at arin.net> wrote:
> On May 16, 2012, at 8:34 AM, William Herrin wrote:
>> The foundation for a claim that preemptive assignment meaningfully
>> harms aggregation is most weak.
>
> Routing aggregation from preemptive provider-independent assignments
> will definitely be much, much less than provider-assigned IPv6 prefixes,
> and while there is some chance of aggregation of successive assignments,
> it is likely that there are very few such end-user successive assignments
> to actually make a meaningful difference.

Hi John,

Respectfully, that's a false comparison. For the folks I'm talking
about, multihomed end users who needed an AS number from ARIN,
provider-assigned IPv6 prefixes are technologically not an option.
Using ISP-assigned /48s for multihoming is widely blocked by  the same
kind of filtering which prevents folks from distinctly routing an IPv4
/32 on the public backbone. As with IPv4 /32 announcements, the
filtering is not universal but it's broad enough to make the use
unworkable.

A correct comparison is between preemptively assigned addresses to AS
holders and addresses only assigned upon request. In that comparison,
increases in disaggregation can only come from two sources:

1. Folks which would not have requested a prefix from ARIN due to an
adequate supply from another RIR *and* choose to make use of the ARIN
prefix just for the heck of it.

2. Folks which would have correctly sized their assignment upon
request, got too small assignment from the preemptive process, need a
larger assignment than sparse allocation can accommodate increasing
the preemptive assignment to *and* didn't realize it until they were
too far into deployment with the preemptive assignment to be willing
to give it up.


#1 is a technicality at best. Aggregation of address assignments is
only important as it relates to aggregation in the routing tables. The
number of orgs in situation #1 who don't otherwise disaggregate their
routes for North America can be expected to be vanishingly small.

#2 requires a whole sequence of improbable events before it results in
any aggregation impact. Again, the number of folks who fall into that
pitfall with preemptive assignment yet could have avoided it with
request-based assignment is likely to be vanishingly small.


Actually, I suppose if I'm being fair I talked broadly about
preemptively assigning addresses broadly to folks holding IPv4
addresses and folks holding AS numbers. ARIN also assigns IPv4
addresses to large single-homed organizations who, unlike multihomed
orgs, face no absolute technological barriers to using ISP addresses.

Does your view on aggregation change if discussion is limited to folks
holding AS numbers because they needed to multihome?

Regards,
Bill Herrin


-- 
William D. Herrin ................ herrin at dirtside.com  bill at herrin.us
3005 Crane Dr. ...................... Web: <http://bill.herrin.us/>
Falls Church, VA 22042-3004