[arin-ppml] IPv6 Transfers (was :Draft Policy ARIN-2018-1: Allow Inter-regional ASN Transfers

[Roberts, Orin]

You could, but then IPv6 routing/fragmentation becomes an issue.

Unless when an ASN is transferred from ARIN all IP networks associated to that ASN are revoked/removed/deleted from ARIN.
ie. I can acquire an ASN that currently exists at ARIN minus any associated IP networks, move it to APNIC/RIPE, then associate IP networks from APNIC/RIPE.

~the same for the reverse.

A proviso would then be, only a clean(ed) ASN can be transferred in/out.

Orin Roberts

From: ARIN-PPML [mailto:arin-ppml-bounces at arin.net] On Behalf Of David Farmer
Sent: February-01-18 1:03 PM
To: Job Snijders <job at ntt.net>
Cc: arin-ppml at arin.net
Subject: Re: [arin-ppml] IPv6 Transfers (was :Draft Policy ARIN-2018-1: Allow Inter-regional ASN Transfers

First, let's move IPv6 transfers out of the ASN transfers thread.

On Thu, Feb 1, 2018 at 11:40 AM, Job Snijders <job at ntt.net<mailto:job at ntt.net>> wrote:
On Thu, Feb 01, 2018 at 12:30:31PM -0500, hostmaster at uneedus.com<mailto:hostmaster at uneedus.com> wrote:
> I would be opposed to allowing inter regional IPv6 Transfers.
>
> One of the main benefits of IPv6 over IPv4 is the reduction of routing
> table size.  Allowing inter regional transfers would start the road to
> larger routing tables.

I'd appreciate evidence that allowing interregional transfers leads to
larger routing tables. Administrative resource management is somewhat
orthogonal to BGP announcements. Whether the resource is managed by RIR
A vs RIR B bears no direct relation to the BGP announcements and routing
tables.

I agree, Inter-RIR transfers doesn't itself imply that the routing table will grow. However, the high level allocations from IANA to the RIRs which are fairly clean in IPv6 today will become fragmented, and likely seriously fragmented if their is signifigant inter-RIR transfers of IPv6. By itself this isn't necessarily a problem, however, IPv6 allocations and assignments have been made in a way to allow most of them to be enlarged in place.  If an allocation is transfered this is no longer easily possilbe to expand the alloation in place.

> We allowed a lot of this in IPv4 because of shortages of addresses.
> This is not in fact true in the IPv6 world. Growth in address use in
> IPv4 resulted in most networks having more than one block of
> addresses.  From what I understand, sparse assigment methods are being
> used in IPv6, allowing those few networks that actually had to grow
> beyond their original allocation to grow into blocks of space right
> next to the space they already occupy, helping to keep the routing
> tables smaller.  During the time we were discussing 2017-5, I asked
> how may ARIN members had grown beyond their original block of IPv6
> addresses, and I believe the answer was zero.

It is by no means zero, I know of seveal allocations that have been expanded.

> IPv6 allows for a host to use more than one address and network.  This
> makes multihoming or renumbering a lot simpler than it was in the IPv4
> world.  I can simply provide more than one router and associated
> network block for each provider, and allow the hosts to obtain an
> address on each of them and to route between them as they see fit.  I
> can also deprecate one of the available networks, and all new
> connections will be made using the remaining networks and routes.
> This allows easy renumbering.
>
> It is not a big hardship to renumber in IPv6 unlike IPv4, so I would
> like to not end up with lots of exceptions in the routing tables, and
> to keep the registration records simpler.

You are describing a very specific deployment model. We cannot assume
that every deployment uses that model, nor build policy based on that
assumption. My own experience tells me that renumbering IPv6 is as much
work as renumbering IPv4.

I also have to agree, the work involed in renumbering is very similar between IPv6 and IPv4.  The diffrence is IPv6 has explicitly condiered renumbering and it is possilbe to renumber IPv6 without a flag day on the local subnet. Whereas with IPv4 each subnet requires a flag day to change from the old to the new addressing.

So I would charterize the diffrence in renumbering in IPv6 verses IPv4, as the impact on an operational network is less with renumber in IPv6, its a far more graceful change with IPv6, but the sheer amount of operational work is comparable between renumbering in IPv6 and IPv4.

Kind regards,

Job

--
===============================================
David Farmer               Email:farmer at umn.edu<mailto:Email%3Afarmer at umn.edu>
Networking & Telecommunication Services
Office of Information Technology
University of Minnesota
2218 University Ave SE        Phone: 612-626-0815
Minneapolis, MN 55414-3029   Cell: 612-812-9952
===============================================
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.arin.net/pipermail/arin-ppml/attachments/20180201/07ad395e/attachment.htm>