Skip to main content

Minutes IETF124: srv6ops
minutes-124-srv6ops-01

Meeting Minutes SRv6 Operations (srv6ops) WG
Date and time 2025-11-04 14:30
Title Minutes IETF124: srv6ops
State Active
Other versions markdown
Last updated 2025-11-05

minutes-124-srv6ops-01

IETF 124 SRv6OPS WG Agenda

Tuesday, 4th Nov 2025, Session I
09:30 - 11:00 (Montreal, Viger)
14:30 - 16:00 (UTC)

Introduction

1.1. Administrivia, Agenda Bashing (Chairs, 5 min) [5/90]

Operator Presentations

2.1 Move to the Host (Nicklous Morris, Verizon)

  • Joel Halpern: Are you assuming compute, edge, core under same administration?
  • Nick: No
  • Joel: That is violation of the RFC
  • Nick: Is everything under the same trust domain?
  • Nick: SRv6 domain in montreal and vancouver can connect over the Internet/secure-tunnel
  • Joel: As co-chair of SPRING, if you want to make a change you need to go to other WG
  • Joel: What type of Host?Conventional, VM?
  • Nick: This is generic - PNF, KVM, basically encap/decap starts at compute infrastructure
  • Joel: Please be claer on trust model and boundries for this to be useful
  • Dan Voyer: This is same operator with multiple domains
  • Jacob: Note the difference between operational domain v/s trust domain
    back to slides
  • Dan: How many processes between groups can you eliminate?
  • Nick: It decreases workload between groups
  • Tom Hill: This could be one trusted domain. This is not about SRv6 domain alone but blast radius - major failure from automation can have huge impact? How to mitigate?
  • Nick: This does not create any more/less wrt to blast radius - single protocol v/s multiple
  • Tom: its about control, PCE controller accross domains, you are putting too many eggs in 1 basket.
  • Nick: This is geo-redudanct PCE, flex-algo on the boxes, there is still logical seperation (segmentation).
  • Dhruv: Share where you are in this journey, what scale do you have?

2.2 SRv6 as a Catalyst for Network Shift (Daniel Bernier, Bell Canada)

  • Jonathan: Selium uses goBGP, do you install routes in kernal tables?
  • Dan: A bgp speaker that talks to the rest of the world, use internals of selium to deploy policies vis eBPF.
  • Dhruv: What is the impact on security and isolation of the workloads?
  • Dan: microsecurity inside ebpf on app level, SRv6 is irrelevent
  • Ron Bonica: Underlay does TE, upper layer solves services problem, you only see End.SID
  • Dan: Multiple locator, anycast SID for service and do TE for it.

Dhruv: Provide feedback to chairs on usefullness of operator presentations. Please spread the word around...

Internet-Draft Presentations

3.1 Deployment Options (Michael McBride, 10 mins) [50/90]
draft-ietf-srv6ops-srv6-deployment

  • Dhruv: SITN - keep the term, but explain on first use; keep I-D focus on deployment options (as an individual); What is missing - what is the strategy to figure out which part of the network to move to SRv6 and why?
  • Tom: As written it feels like it says SR-MPLS is a natural progression to SRv6 and that is not the case. Be very factual. Document - Can co-existence benefit or should use 6PE? Make it clear that SR-MPLS is not a stepping stone to SRv6.
  • Gyan: Should roll in term SITN to dual plane.
  • Nick: Term SITN is okay; Dont skip VXLAN migration

from chat

Tom Hill
@mike The security considerations section needs a reference to the security considerations draft. The reason that's being written is because the security considerations sections of the existing RFCs weren't comprehensive enough. :) draft-ietf-spring-srv6-security
Nat Kao
@Mike Maybe we can discuss more about the inter-AS VPN/CSC migration cases on the list. I think the use case from Daniel's presentation can be a good reference.
Mike McBride
Agree Nat, inter-as vpn migration, and now vxland migration, will likely be the two major additions for the next rev. thanks. Will ping you, Daniel and the list.
Daniel Bernier
as per your comment @tom hill ... I agree, some live and breathe by VxLAN, GENEVE or MPLS and are quite happy with what it provides them. but operators living in the world of passing bits at commodity revenues either are happy being and try the game of supporting any/every flavors of SASE/SDWAN with their pitfalls and shortcomings OR they can try and evolve themselves.

3.2 Protection (Yisong Liu/Yao Liu, 10 mins) [60/90]
draft-liu-srv6ops-sr-protection

  • Dhruv: Much of this is common between SRv6 and SR-MPLS, should this belong to SRv6OPS?
  • Yisong: Protection is very similar and share the same policy
  • Dhruv: Need to discuss with SPRING chair on what is the best way forward for this I-D
  • Rakesh Gandhi: (slide 4) Liveness check, good to see STAMP - you get latency, loss and liveness.

3.3 OAM Deployment (Liyan Gong, 10 mins) [70/90]
draft-gong-srv6ops-srv6-oam-deployment

  • Zafar Ali: You can ping BSID; in different domain if you dont have reverse path - that could be an issue, but it can be clarified
  • Rakesh: Encap mode - other drafts in SPRING, better to align terminology. (Path ECMP), with flow label you dont know which segment-list path, that should be expanded.

from chat

Dhruv Dhody
One meta comment on OAM draft - the I-D title deployment considerations but it is more a list of OAM options, what i would like is trade-offs between approaches or operational guidance on when to use what

3.4 SRv6 policy selection (Feng Yang, 10 mins) [80/90]
draft-yang-srv6ops-policy-selector

  • Zafar: This is not SRv6 specific, there is also composite draft.

from chat

Dhruv Dhody
What is the relationship between this policy selector I-D and Intelligent Routing https://datatracker.ietf.org/doc/draft-yang-srv6ops-intelligent-routing/

Dhruv Dhody
Also, is this completely internal to a node, what needs to be standardized?

Ketan Talaulikar
Why use composite sr policy if different services want to use different underlying sr policies? Why not directly use those underlying sr policies? Am I missing something?

3.5 Addressing Recommendations (Jakub Horn, 10 mins) [90/90]
draft-horn-srv6ops-srv6addressing

  • Nick: 5f00 does not have the same property as ULA, be careful with that framing
  • Jeff: From DC space, this is harmful. F1616 may be better and worth evaluation.
  • Tom: This is only for NEXT-CSID. There are other flavors.
  • Dhruv: Addressing without compression should also be considered.
  • Jakub: Current focus on most common. Can add para for other formats

Last updated by Dhruv on 5th Nov 1 PM