SPRING @IETF-127

Monday 16 November 2026
Room: Continental 5
16:30 - 18:30, Local
Log into the IETF datatracker to access:

SPRING WG Meeting Agenda

16:30 SPRING Status - Chairs (10 mins)

draft-ietf-spring-cs-sr-policy is blocked by four other drafts we have
not finished.
Zafar: draft-ietf-spring-sr-policy-yang is undergoing some changes and
being moved forward.
Joel: For the authors of other drafts, please be aware of this situation
and move your draft forward.

16:40 ICMP Error Handling for VPNs in SRv6 Networks (10 mins)

Presenter: Zafar Ali
draft-alivar-6man-icmp-error-srv6-vpn

Authors would like to receive WG feedback and the WG to adopt the
document.
No comments.

Joel: We will take this to the list. For the authors, please get
engagement on the list before we do adoption.

16:50 Flexible Candidate Path Selection of SR Policy (10 mins)

Presenter: Yisong Liu/Changwang Lin
draft-ietf-spring-sr-policy-flexible-cp-selection

Greg Mirsky: It seems that you want to use STAMP for hop-by-hop
measurement. But STAMP is for end-to-end performance measurement not for
Hop by hop performance measurement. Maybe iOAM or alternate marking can
be used.

Yisong: We have included these perfomrance measurement methods. We can
discuss more offline.

Rakesh: Similar comment to Greg. Destination option is processed by
desination not hop by hop. It is also not clear whether you are
measuring the dataplane bandwidth or control plane bandwidth for the
interface. It needs more work there.

17:00 Signaling RSVP-TE Tunnels on an SRv6 Forwarding Plane Using End.X Segment Identifiers

Signaling RSVP-TE Tunnels on an SR-MPLS Forwarding Plane Using Adjacency Segment Identifiers (15 mins)

Presenter: Vishnu Pavan Beeram
draft-beeram-mpls-rsvp-sr
draft-beeram-teas-rsvp-srv6

Sue: Is it dataplane or control plane or both? If combination, is this
making assumptions on the candidate path wandering through. I don't see
how it's all going to work. Are you assuming composite in this?

Vishnu: We are adding a knob in the data model, which says bandwidth
management or RSVP signalling some knob which triggers this signalling.
Once the knob is enabled, we are leveraging RSVP signalling, making
bandwidth reservation everywhere along the path. No change on the
dataplane. Only RSVP control plane is brought here. No BGP.

Joel: It's RSVP-TE from the headend carrying the control signalling as
per RSVP-TE as always. No BGP.

Sue: That is not my problem then.

Zafar: How does RSVP facility protection work when you have protected
adjacency sid.

Vishnu: There are MPLS documents that talks about a couple of options.
You could just use TLFA to protect a signal primary path. The only
extension needed is how you ensure the primary path from not going down
when a failure happens. In terms of signalling using a facility backup,
there is no way to say that a protection sid is catered by a specific
mechanism. That is a gap in IGP advertisement which we will fix in a
subsequent document.

Zafar: These are not in the draft, neither the RSVP-TE signalling part.

Vishnu: There is a section there. The only gap is that in the RSVP
protocol because once the link goes down and if you are using a
non-signaling backup you need to find ways to preserve the primary path
from going down. The node SID can be used to fix it. That is an update
to RFC4090, and we have got that covered.

Zafar: These need to be specified for the review. How is RSVP using
IGP-ownded label?

Vishnu: Please read the draft. Nothing is being allocated by RSVP. The
lables are IGP owned. If you want to use RSVP allocated labels just use
RFC8577.

Zafar: If IGP changes the labels what happens?

Vishnu: We have the signalling plane. Once the IGP changes the node
would know, and there will be a signal. There will be no micro loops.

Zafar: There are too many pieces right now. We need to have a draft in
spring to cover them.

Vishnu: Please send to the list. Happy address.

Greg: The use of IGP-TE is implied here? Please do mention it
explicitly.

Tarek (Cisco): We cannot do bandwidth reservation over all SR paths or
specifically those are not strict? The SR paths are not all strict.

Vishnu: We are clear that we only use adjacency sids because the
reservations are made per TE link. For the longish path, we also have
the solution.

Jie: The WG needs to first evaluated whether the distributed bandwidth
management is needed for SR. Then we can evaluate all the approaches.

Vishnu: RSVP-TE is the widely deployed distributed bandwidth management
solution. If there is any other solution please send to the list.

Joel: If there are other solutions please bring that to the list.

Jie: Integrating the RSVP control plane with SR data plane may introduce
architecture changes and additional functionality and control protocols
will be introduced.

Vishnu: No violation.

Andrew: There is this assumption where an SR Adjacency sid is only used
by sr tunnels or sr native traffic. By bringing RSVP tunnel that now
uses those adjacency sids breaks this subtle assumption. The WG needs to
first discuss more.

Vishnu: I have added an operation consideration section on this.

17:15 BGP SR Policy Extensions for BFD Configuration (10 mins)

Presenter: Song Liu
draft-li-idr-bgp-sr-policy-bfd-extension

Jeff: The spring wg needs to focus on is where these BFD things go
because for the candidate paths we have more than one way that we can
represent where this stuff is getting glued as attributes into the
various state that we are passing around in control plane protocols.
What BFD is actually protecting in such circumstances. When you do BFD,
talk about what that means for a given use case.

Song: Will discuss in the mailing list.

17:25 Distributed Bandwidth Management for Segment Routing (10 mins)

Presenter: Shraddha Hegde
draft-hegde-spring-sr-distributed-bw

Zafar: This work could be progressed more in spring so we can have basis
for other protocol work. You said that OAM doesnot need any change, but
we don't know if the SR data plane needs changes for protection or not.
This needs to be developed in spring. Based on this work, other work can
be progressed.

Shraddha: My understanding is that it will not change anything. But we
will include protection in the draft and bring to the WG.

xx: The traffic engineering for RSVP and SR are separated. Now you are
mixing signalling and traffic engineering.

Shraddha: Your concern is about the link attributes that will be used
for this mechanism will be based on application specific attribute or it
is going to be the legacy attributes. In this work, there is SR
forwarding plane and RSVP control plane. It needs to be more clearly
specified. This is the right document to talk about it.

17:35 SRv6 for Multipath Traffic Engineering (10 mins)

Presenter: Abhishek Chakraborty, Jayant Kumar Agarwal, Vishnu Pavan
Beeram, Siva Sivabalan
draft-ajp-spring-srv6-mpte

Andrew: In this case, we are taking the same locator and putting it on
every node in the DAG. Is it fair to say that effectively because it's
the same address that's being programmed everywhere? It is the
equivalent of like a state route with an IPv6 to multiple hops on every
node.

Abhishek: In SRv6, locator identifies a reachability to a node. This is
the same for tunnels. It is realising this particular tunnel with this
particular locator which has a forwarding semantics for unidirectional
tunnel.

Andrew: The locator is just a forwarding instruction.

Abhishek: The locator is a static route on each of these nodes to reach.
A locator is just a forwarding semantics for any identified node or a
tunnel.

Greg: You mentioned to use SBFD to monitor this policy. How does the
split paths from T2 each individually monitored using one SBFD session?

Abhishek: We are not inventing any new solutions here.

Greg: SBFD is only applicable only to point to point. Here you have a
point to multipoint. How point-to-point is mapped to point-to multipoint
environment?

Abhishek: Each junction to junction can run its own SBFD over ECMP over
its links and can realizes its own SBFD states.

Greg: So you have to generate twice the SBFD packets. You are making
claim which is not technically substantial.

if time allows

17:45 Framework for Time-Scheduled Color based SR Policy Selection (10 mins)

Presenter: Cancan Huang
draft-huang-spring-time-scheduled-color-framework

Li Zhang: Why do you use the schedule at policy level instead of doing
schedule at candidate pass or segment list level?

Cancan Huang: The canadidate path just serves for a dedicated color, the
path itself, not bandwidth or latency. The policy can present the
bandwidth.

17:55 SR-TE Path Based on Quality-Aware Dynamic Adaptive Adjustment (10 mins)

Presenter: Tingting Han
draft-cheng-spring-srpolicy-ar-framework

Joel: We are out of time. Please take discussions to the list.

18:05 YANG Data Model for SR Policy Group (5 mins)

Presenter: Changwang Lin
draft-gong-spring-sr-policy-group-yang

Speaker Shuffling Time: 10 minutes
Total Presentation Time: 110 minutes