TSVWG

Summary

The Transport Area Working Group (TSVWG) met at IETF 126 to discuss several active working group drafts, upcoming specifications, and experimental proposals. Key areas of focus included user port experiments, SCTP DTLS encapsulation and key management, the FQ-PIE AQM algorithm, L4S operational guidance, WebRTC data channel acceleration, and mobile L4S deployment challenges.

An outgoing Liaison Statement (LS) was sent to IEEE 802.11 and the Wi-Fi Alliance (WFA) regarding the publication of RFC 9956 (which updates RFC 8325).

Key Discussion Points

1. User Ports and Port Identifiers for Experiments

Presenter: Wes Eddy
Slide Deck: User Ports and Port Identifiers for Experiments
Discussion:
- The draft has returned to the working group after receiving five "Discuss" ballots from the IESG. It is currently expired.
- The primary concerns from the IESG relate to firewall traversal, security implications of the new ports, and the mechanism for Port Experiment Identifiers (PEXIDs)—specifically, how they operate across different transport protocols.
- Gorry Fairhurst suggested that adding concrete examples to the draft would help clarify the intended use cases to the IESG.
- Zaheduzzaman Sarker and David Black suggested adding Wes Eddy as a co-editor and utilizing a port reviewer as a draft shepherd to help guide the document through the re-balloting process. Michael Tuexen and Mike Heard offered to assist with reviews and text.

2. DTLS Chunk for SCTP

Presenter: Magnus Westerlund
Draft: draft-ietf-tsvwg-sctp-dtls-chunk
Slide Deck: DTLS Chunk for SCTP
Discussion:
- Recent updates include choosing a single DTLS 1.3 record format configuration (no length fields, 16-bit sequence number) and removing the indicator bits in the chunk header. Support for simultaneous open was added using a tiebreaker field.
- Interoperability has been successfully demonstrated between two independent implementations: Red Hat (Linux kernel SCTP) and Ericsson (user space stack using WolfSSL).
- A remaining minor issue is whether to retain both the abstract and socket API descriptions in the main body. Gorry Fairhurst noted that APIs are typically informative appendices, but Magnus Westerlund explained they help clarify the assumptions between the key management layer and the chunk layer.
- Michael Tuexen requested early IANA allocation of error causes to facilitate integration into open-source releases.

3. Key Management for DTLS Chunk

Presenter: Magnus Westerlund
Drafts: draft-ietf-tsvwg-dtls-chunk-key-management & draft-porfiri-tsvwg-sctp-dtls-handshake
Slide Deck: Ericsson's Proposal for Key Management for DTLS Chunk
Discussion:
- Magnus Westerlund presented an alternative key management proposal (draft-porfiri-tsvwg-sctp-dtls-handshake) that switches from DTLS to TLS 1.3. This approach leverages standard TLS stacks without requiring extended key updates, reducing implementation complexity.
- Michael Tuexen supported adopting the new TLS 1.3-based proposal as the primary working group path.
- Zaheduzzaman Sarker suggested evaluating both approaches further on the mailing list before officially overturning the design team's original decision.

Session Poll 1

Question: I have read/reviewed some version of draft-porfiri-tsvwg-sctp-dtls-handshake
Results: Yes: 5, No: 43, No Opinion: 1 (Total: 81)

4. Flow Queue PIE (FQ-PIE)

Presenter: Mohit P. Tahiliani
Draft: draft-ietf-tsvwg-fq-pie
Slide Deck: FQ-PIE
Discussion:
- FQ-PIE combines flow queuing with the PIE AQM algorithm. It has been tested in Linux, FreeBSD, and NS3.
- Recent hackathon testing focused on Wi-Fi and mobile hotspots. Results showed that FQ-PIE exhibits better 90th percentile tail latency compared to FQ-Codel because it drops packets at enqueue time rather than dequeue time.
- A patch implementing per-flow statistics has been submitted to the Linux net-next branch.
- Chris Box asked if simulation results were consistent across non-Wi-Fi environments. Mohit P. Tahiliani confirmed that published NS3 evaluations show consistent benefits. Greg White noted that CableLabs observed similar benefits when comparing PIE to Codel for DOCSIS AQM.

5. L4S Operational Guidance & Static Rate Management (SRM)

Presenter: Greg White
Draft: draft-ietf-tsvwg-l4sops
Slide Deck: L4Sops Draft and L4S Interops
Discussion:
- A new "Static Rate Management" (SRM) mechanism using a Two-Rate Three-Color Marker (TRTCM) has been proposed as a separate draft by Kuhn de Schepper and Miroslav Kovac.
- Greg White updated draft-ietf-tsvwg-l4sops (draft-10) to include a brief reference to this new SRM mechanism.
- Martin Duke raised concerns about whether general L4S mechanisms like SRM fit within the scope of the l4sops draft, which is specifically tasked with addressing RFC 3168 coexistence.
- Miroslav Kovac clarified that SRM is highly beneficial for high-speed edge devices shaping to subscriber line rates, as it reduces the queuing delay needed to absorb bursts.
- The WG discussed whether to pause the publication of the l4sops draft to allow the SRM draft to mature.

Session Poll 2

Question: Should we park l4sops until the SRM draft can be considered?
Results: Yes: 8, No: 14, No Opinion: 16 (Total: 84)

6. SCTP Negotiation Acceleration Protocol (SNAP) for WebRTC

Presenter: Philipp Hancke
Slide Deck: SCTP Negotiation Acceleration Protocol (for WebRTC)
Discussion:
- Establishing WebRTC data channels currently requires six RTTs. The SNAP proposal piggybacks the SCTP INIT chunk inside the SDP exchange (using base64), eliminating the on-the-wire SCTP handshake and reducing connection time by two RTTs.
- OpenAI deployed this in their mobile apps, verifying a 2 RTT P50 improvement.
- Michael Tuexen suggested that rather than base64 encoding the entire INIT chunk, the SDP should explicitly declare the minimal necessary parameters (such as RWND, stream counts, and supported extensions) to avoid unnecessary complexity and network-byte-order conversions.
- Martin Thomson questioned the utility of modifying SCTP, suggesting that WebTransport is a cleaner, single-RTT solution. Philipp Hancke noted that migrating large, existing WebRTC deployments to WebTransport remains highly complex.
- Gorry Fairhurst recommended continuing the technical discussion on the TSVWG mailing list before deciding which working group (TSVWG or AVTCORE) should adopt the work.

7. Mobile L4S

Presenter: Sebastian Premba
Slide Deck: mobile-l4s
Discussion:
- This new draft addresses challenges when deploying L4S in mobile ecosystems, including path blackholing/filtering, link-layer queue management, and the risk of applications misusing ECT(1) markings for "priority" treatment.
- Stuart Cheshire emphasized that L4S is not a priority mechanism; it is an informational mechanism to help clients avoid self-congestion. He noted that queue protection (e.g., dual-queue AQM) is critical to discourage abusive flows.
- Zaheduzzaman Sarker questioned whether the draft should be framed as mobile-specific or generalized for the wider Internet, pointing out that 3GPP has its own standards processes.
- Lorenzo Colitti replied that device manufacturers and OS teams face extreme confusion when implementing L4S, and this draft serves as a vital checklist of requirements across the network, OS, and modem.

8. Non-Presented Drafts

Due to time constraints, the following presentations were deferred to the mailing list:

Export of ECN Information in IPFIX (Slide Deck: Export of ECN Information in IPFIX)
TCP and UDP checksum calculations for ILNP (Slide Deck: TCP and UDP checksum calculations for ILNP)

Decisions and Action Items

User Ports Experiment: Wes Eddy will be added as a co-editor to draft-ietf-tsvwg-user-ports-and-port-identifiers-for-experiments to address the IESG discusses. The team will work through concrete firewall and PEXID examples to prepare for a future re-ballot.
SCTP DTLS Chunk: The authors of draft-ietf-tsvwg-sctp-dtls-chunk will resolve the minor API location issue.
Early IANA Allocation: The chairs will initiate early IANA allocations for the error causes in draft-ietf-tsvwg-sctp-dtls-chunk.
L4Sops Draft: Due to the split poll results on whether to park draft-ietf-tsvwg-l4sops for the SRM draft, the chairs will take this question to the mailing list to determine consensus.

Next Steps

DTLS Chunk WGLC: Once the API formatting is resolved, the chairs expect to initiate Working Group Last Call for draft-ietf-tsvwg-sctp-dtls-chunk before the next meeting (IETF 127).
Key Management Evaluation: Working group members are encouraged to review both draft-ietf-tsvwg-dtls-chunk-key-management and the new TLS 1.3 proposal (draft-porfiri-tsvwg-sctp-dtls-handshake) ahead of a potential adoption call at or near IETF 127.
FQ-PIE WGLC: The chairs plan to issue a WGLC for draft-ietf-tsvwg-fq-pie in the near term. Any outstanding issues should be filed on the project's GitHub repository.
SNAP and Mobile L4S: Technical discussions on the SNAP WebRTC optimization and the mobile-l4s deployment draft will continue actively on the TSVWG mailing list.

Automatic IETF Minutes