Markdown Version | Transcript | Session Recording | Session Materials
Session Date/Time: 20 Mar 2026 01:00
TSVWG
Summary
The Transport Area Working Group (TSVWG) met at IETF 125 in Shenzhen to discuss a variety of transport-related specifications and operational updates. Key topics included the security of SCTP via DTLS, refinements to the FQ-PIE algorithm, new IPFIX elements for ECN monitoring, and large-scale operational reports on L4S deployment. Significant progress was reported on the DTLS chunk for SCTP, which is nearing Working Group Last Call (WGLC).
Key Discussion Points
1. Chairs' Introduction and Agenda Bashing
Speakers: Zaheduzzaman Sarker, Martin Duke
- Slides: Chairs slides
- One RFC was recently published, and several others are in the RFC Editor queue.
- Milestones: Chairs adjusted milestone dates to provide better predictability for other Standards Development Organizations (SDOs).
- Document Review: A call was made for more cross-sector reviews. draft-ietf-tsvwg-l4sops is currently in Last Call.
- TSV Review Team: Participants were encouraged to join the Transport Area Review Team to assist with IESG reviews.
2. DTLS Chunk for SCTP
Speaker: Magnus Westerlund
- Slides: DTLS Chunk for SCTP
- Drafts: draft-ietf-tsvwg-sctp-dtls-chunk, draft-ietf-tsvwg-dtls-chunk-key-management
- Design Update: The design team decided to remove the DTLS Connection ID (CID). SCTP already provides robust demultiplexing via V-tags and port/address pairs, making the CID redundant and adding unnecessary complexity to the API.
- ASCONF Updates: The draft now updates RFC 5061 to specify that the DTLS chunk (rather than SCTP-AUTH) should be used for authenticating address reconfiguration messages.
- Interoperability: Implementation work is ongoing in FreeBSD and Linux. No major technical hurdles remain.
- Key Management: Work on draft-ietf-tsvwg-dtls-chunk-key-management is trailing the main chunk draft to ensure API alignment.
3. FQ-PIE Refinements
Speaker: Mohit P. Tahiliani
- Slides: FQ-PIE
- Draft: draft-ietf-tsvwg-fq-pie
- Technical Detail: The draft recommends using the timestamps approach for calculating queue delay rather than Little’s Law.
- L4S Support: Discussion continues on adding L4S support (CE threshold). Mohit P. Tahiliani suggested that a separate draft might be appropriate to define L4S support across all FQ-based mechanisms (FQ-CoDel, FQ-PIE, FQ-Cobalt).
- Implementations: FQ-PIE is now supported in the Linux kernel (OpenWrt), FreeBSD, and the ns-3 simulator. Recent work integrated FQ-PIE into
go-tcandQOSMate.
4. Export of ECN Information in IPFIX
Speaker: Shueyan
- Slides: Export of ECN information in IPFIX
- This individual draft proposes new IPFIX Information Elements (IEs) to monitor L4S traffic (ECT(1), ECT(0), CE, and non-ECT) across IPv4, IPv6, and MPLS.
- Discussion: Gorry Fairhurst questioned why only the 2 ECN bits were being exported instead of the full 8-bit Traffic Class/DSCP field. Shueyan noted that while DSCP IEs exist, specific ECN statistical IEs are needed for L4S visibility.
- Operator Feedback: Jason Livingood expressed strong support, noting that the lack of standardized IPFIX ECN reporting is currently an operational gap for ISPs deploying L4S.
5. HP-WAN Hackathon Report
Speaker: Daniel Huang
- Slides: HP-WAN Hackathon report
- Results from a prototype of High-Performance Wide Area Networking (HP-WAN) were presented. The project used RSVP-based signaling for host-to-network coordination to achieve high throughput and low latency.
- Collaboration: Daniel King noted that the hackathon highlighted a need for a consistent "workload model" and a feedback loop from the network to the orchestrator regarding congestion state.
6. L4S Operational Update
Speaker: Jason Livingood
- Slides: L4S Update - Livingood
- Deployment: Comcast has enabled L4S and NQB for over 10 million homes (approx. 350 million devices). Key use cases include cloud gaming (NVIDIA GeForce NOW, Valve Steam) and real-time communications (FaceTime).
- Technical Challenges:
- FEC Interactions: Jason Livingood noted that existing Forward Error Correction (FEC) algorithms often have baked-in delay assumptions that conflict with the ultra-low latency provided by L4S.
- ECN Bleaching: Stuart Cheshire (Apple) reported instances of ECN bleaching or incorrect CE marking (e.g., in certain mobile networks) which causes throughput degradation.
- Client Status: Stuart Cheshire confirmed that Apple products support Prague (as a modification to Cubic/QUIC) and are currently performing A/B testing in the field.
Decisions and Action Items
- Decision: The design team for draft-ietf-tsvwg-sctp-dtls-chunk confirmed the removal of the DTLS Connection ID from the specification.
- Action: Participants are requested to review draft-ietf-tsvwg-l4sops before the end of the Last Call period.
- Action: Magnus Westerlund to perform an editorial pass on the SCTP DTLS Chunk draft to prepare for WGLC.
- Action: Jonathan Lennox noted that AVTCORE drafts regarding DTLS in STUN and SCTP in SDP should be cross-posted/presented in TSVWG due to shared interest.
Next Steps
- WGLC Target: The chairs and authors aim to issue a Working Group Last Call for draft-ietf-tsvwg-sctp-dtls-chunk prior to the IETF 126 meeting in Vienna.
- L4S Operations: The working group will continue to monitor field results from the Comcast and Apple deployments to refine operational guidance.
Related Documents
draft-ietf-tsvwg-dtls-chunk-key-management, draft-ietf-tsvwg-fq-pie, draft-ietf-tsvwg-l4sops, draft-ietf-tsvwg-sctp-dtls-chunk