Interim Session - 20 February 2023
WG co-Chairs: Erik Bais, James Kennedy, Leo Vegoda
On 20 February 2023 from 14:00 to 15:00 (UTC+2), the Address Policy Working Group held a remote session via Zoom to review the RIPE IPv6 policies and identify desired policy outcomes and set priorities. More background on this work is published on RIPE Labs.
Recording
Minutes
Co-chairs: Erik Bais, James Kennedy, Leo Vegoda
Scribe: Anastasiya Pak
1. Introduction
Leo Vegoda welcomed attendees and reminded everyone to follow the Code of Conduct.
The agenda for the meeting was to discuss the following issues relating to IPv6:
- Unclear guidance on documentation requirements for large allocations
- Complex language and implementation of the HD-ratio
- IPv6 space not assigned/allocated on a nibble boundary
- Policy limitations for PI assignments
2. Nibble boundaries
The participants discussed a desired policy outcome for nibble boundaries.
Erik Bais referenced a comment by Tina Morris, AWS, at RIPE 85 about the issues certain companies were facing because of the micro allocations.
Sander Steffan said that addresses were more readable if you stuck to nibble boundaries.
Jan Žorž agreed that nibble boundaries wouldn’t create any harm and reminded the group of the reasoning behind the /29 policy, and the reservation of a /26 by the RIPE NCC for every /29.
Jordi Palet Martínez suggested that LIRs which met the initial allocation criteria were eligible to receive an initial allocation of /32 up to /28 without needing to supply any additional information. Jordi asked if they wanted to change something in the policy about existing allocations.
Sander said this could be problematic as some couldn’t grow to a /28.
Matt Parker, RIPE NCC, suggested they ask for an impact analysis from the RIPE NCC before putting something into the policy about expansion because it might be technically impossible.
The group agreed.
Jan asked if the suggestion for the RIPE NCC would be to reserve the space up to the next nibble boundary for every allocation. The participants agreed.
Sander reminded everyone to not forget about the PI space, and if some users would then automatically grow from /48 to /44.
Erik mentioned that PI holders had different documentation requirements to receive additional allocations.
The group agreed that the initial and subsequent allocations/PI assignments should always be on the nibble boundary and to check whether this always doubled the allocation.
3. HD-Ratio
Sander suggested they get rid of the HD-Ratio and find an alternative. He said it was important to provide clear guidelines for the RIPE NCC to follow for fairness and education purposes.
Raymond Jetten added that the formula was complicated.
The participants agreed that a tool which matched the way network engineers planned and thought was needed.
4. Documentation
Participants moved on to discuss documentation requirements for large allocations and related issues. Erik brought up the case of Rinse Kloek’s company, which had struggled with documentation during the rollout of IPv6 in their FTTH network.
Rinse said they made a plan to merge several /29s that were acquired via M&A into a single /26 to support regional aggregation. The RIPE NCC could not do this as the policy does not support this kind of cleanup. The consequence is that their customers each got a /56 instead of a /48.
Maximilian Emig suggested they always offer a contagious allocation, even if a larger allocation was requested for routing reasons.
Erik thanked Maximilian for his comment and said this was relevant to the nibble boundary discussion.
Jan remembered the old days when he asked a RIPE Meeting audience if they understood the HD-Ratio. It was still not well understood, which meant it was not serving its purpose. A good solution would be counting the number of users or sites with a /48 and making a projection for the next 10 years.
Erik supported this comment and Rinse agreed with both.
Angela Dall'Ara, RIPE NCC, asked if the RIPE NCC always allocated addresses on the nibble boundary, would that mean that they automatically went from /28 to a /24? She also wondered what the RIPE NCC should request as documentation/justification and to which extent the addressing plan should cover the new size.
Sander said that it was still needs-based. He also added that it should be based on the number of ‘customers’, projected over 5-10 years and 1-2 aggregation levels could be added.
Responding to previous points raised in the group, Angela added that, based on the current policy, there was no possibility to swap allocations with different ranges on request.
The participants continued the discussion on documentation requirements, and there was agreement that the requirements should be predictable, simple and clear.
Leo asked what was their concern, given that running out of IPv6 space was not the primary one.
Erik replied that the goal was to have an IPv6 policy that was similar to the AS policy so they didn’t need to touch it every 18 years.
Jordi added that it was important to avoid abuse.
After the discussion, the participants agreed that a desired policy outcome would be the following: published and predictable requirements that were easy to implement; renumbering should not be required to get more addresses (but must be an option).
Finally, the participants agreed that more meetings were needed to discuss PI assignments and other issues related to IPv6 addressing. It was noted that the next meeting would be scheduled soon to continue the discussion.