Re: [Idr] WGLC on draft-ietf-idr-as-private-reservation-00 concluded, extended to consider ASN range

[Jon Mitchell <[e-mail]>]

On Wed, Dec 19, 2012 at 03:41:08PM +0100, Robert Raszuk wrote:
> Hello David,
> 
> > So, if someone can provide a reasonable justification for 100M private use
> > ASNs, or even 10M for that matter, I'm all ears
> 
> I think we are all observing exponential rate of office backends being
> outsourced. Also notice that grow of public AS numbers is not that
> steep. Through last 12 years we allocated 50K !
> http://www.potaroo.net/tools/asns/

I agree that this is the case.

> So IMHO it is not that unrealistic that some global SP (maybe Google,
> Microsoft, Apple2 etc... :) could offer free Internet access
> everywhere they could reach or where law mandates free "last mile".
> Would we want to limit number of their global customers ? Would we
> really need to invent new hacks to work around private AS number
> starvation ? Of course it is dead clear that Internet access does not
> mandate use of BGP even for multihomed sites (example: LISP). But this
> is IDR so I guess talking about BGP here is ok.
> 
> My seemingly wild suggestions to offer half of the space as private or
> think about make AS hierarchical were aiming precisely in such new
> EBGP (or EBGP-lite) peerings for large numbers of IPv4/IPv6/VPNs
> customers.
> 
> If someone would to use this new RFC for such applications let me
> reverse the question .. do we have sufficient reasons to limit the
> space to 1M customers ? Are we afraid that even if we reserve 31 bits
> the public AS will experience shortage anytime soon considering
> current grow rate of public as number assignments ?

I think there is hope that there will be interesting growth on the
Internet or globally unique identifier use cases with Public ASN
registrations and that RIR policies may change to encourage these use
cases over time.

It's hard to tell if you are seriously suggesting a 50% proposal still,
but I would consider this is a substantive change to the intent,
motivation and content of the draft which is for a single organization
(and Private Use) and would probably require a new draft.  I can't
accept a proposal to move to 50% in the draft and there seemed to be no
energy from anyone that this is the right direction.  It also seems to
then remove the possibility that these types of allocations would be
from Public space, and disallow other proposals such as the one you
started a seperate thread about by consuming too much of the overall
space to make them acceptable any longer (there by forcing more use
cases with large number of ASN needs into Private Use space).

However if your overall point is that 10M - 100M is not unreasonable for
this draft as is, I agree and think certain variations of use cases
(more the backend ones) you identified above would be Private Use if a
single organization is allocating and managing the ASNs (in my mind more
likely the large office backend customers than a provider, or a provider
using BGP in these scenarios transparently / not managed by the customer).

Jon