[arin-ppml] Discussion Petition of ARIN-prop-125 Efficient Utilization of IPv4 Requires Dual-Stack

Matthew Kaufman matthew at matthew.at
Thu Jan 6 22:07:22 EST 2011


On 1/6/2011 5:58 PM, Chris Grundemann wrote:
> In response to feedback both on and off list, the other originators of
> this policy and I have agreed on a much lighter weight version of
> prop-125. Although I can not revise the proposal while it is under
> petition, I would like to share the new text with everyone and ask for
> your support. If the petition is successful, I will submit this new
> text as version 2.
I do not support the petition.

I also would not support the proposal or the petition as modified below. 
See comments inline.
> ~Chris
>
> Changes in this version:
>   1) Added parallel IPv6 networks in addition to dual-stack.
>   2) Removed the explicit application to transfers.
>   3) Removed all retro-active requirements.
>
> ## Policy Statement:
>
> * Add the following sections to section 4.1:
>
> 4.1.2. Efficient Utilization
>
> IPv4 addresses are a finite resource and as such are required to be
> efficiently utilized by resource holders in order to maximize their
> benefit to the community.
I don't have a problem with this. We just don't agree on "efficiently 
utilized".
> 4.1.3. IPv6 Deployment
>
> 4.1.3.1 Dual-Stack
>
> Dual-stack refers to configuring both an IPv4 and an IPv6 address or
> network together on the same network infrastructure.
Ok.
> 4.1.3.2 Parallel Deployment of IPv6
>
> Parallel deployment of IPv6 refers to the building out of a second,
> parallel IPv6 network along side of an existing IPv4 network. In order
> to be considered a parallel deployment of IPv6, the IPv6 network
> infrastructure may be physically or logically seperate from the IPv4
> network but must provide the same functionality and services.
This might be vague enough to be acceptable, as long as it is understood 
that "same functionality and services" doesn't necessarily mean 
"identical".

At the most basic level, it is clear that at current growth levels it 
isn't at all necessary to provide as many IPv6-facing servers as 
IPv4-facing servers to get the same response times.

But in some cases the "same functionality" is impossible to provide for 
one or the other protocol, and it wouldn't make sense to try.
> 4.1.3.3 Efficient Utilization of IPv4 Requires Ipv6
>
> All new IPv4 addresses assigned or allocated to an organization must:
>   - be deployed on dual-stacked interfaces along with IPv6 addresses, or
>   - be deployed along with a parallel deployment of IPv6
Again, do you mean "parallel concept" so that as long as you can do 
similar things with IPv6 that counts, or do you mean "strictly parallel" 
in that for every box behind the IPv4 load balancer there must be a box 
behind the IPv6 load balancer?

And what would a "parallel deployment" of IPv6 mean for cases where it 
makes *no sense at all* to try to provide the same thing? (Examples 
include my hypothetical legacy computing museum from last week, or a 
proxy service that is only relevant to one protocol or the other)
> 4.1.3.4 Service Parity
>
> When IPv4 addresses are used to provide an Internet facing service,
> the service must be fully IPv6 accessible.
What do you mean by "fully IPv6 accessible"? Do you mean "the firewall 
lets the same ports through"? Or do you mean "your IPv6 transit must be 
as well-connected as your IPv4 transit"? Does that mean I need a second 
IPv6 provider if mine has a peering disagreement with one of the other 
major ones? Does it mean that I can never be "fully IPv6 accessible" as 
long as there are eyeball networks that are IPv4-only?
> * Add the following sentance to the end of sections 4.2.1.6,
> 4.2.2.1.2, 4.2.2.2.1, 4.2.3.1. and 4.3.4:
>
> In accordance with section 4.1.3, all new addresses must be deployed
> with IPv6 addresses and all Internet facing services provided by new
> addresses must be fully IPv6 accessible.
And again, doesn't make sense. If my "Internet facing service" is 
"telnet access to NeXTStation machines" there isn't really an IPv6 
equivalent, is there? And who would expect it anyway?
> * Re-write section 4.2.3.4.1. to:
>
> Reassignment information for prior allocations must show that each
> customer meets the 80% utilization criteria, the IPv6 deployment
> criteria and must be available via SWIP / RWhois prior to your issuing
> them additional space.
Given that we don't know how ARIN would audit this, it isn't clear how 
an ISP would.

And you've covered "downstream reassignment" and "Internet-facing 
services" but there's all sorts of other reasons someone might apply for 
IPv4 addresses... what do they need to do in order to "pass the test"?

Matthew Kaufman




More information about the ARIN-PPML mailing list