**Session Date/Time:** 10 Nov 2022 13:00

# webtrans

## Summary

This webtrans meeting at IETF 115 covered several key topics, including updates from the W3C on the web transport API, discussion of the Capsule Design Team's work, the negotiation of web transport, handling stream resets, and related HTTP/3 issues. Decisions were deferred on some topics pending further discussion on the mailing list and in other working groups.

## Key Discussion Points

*   **W3C Web Transport API:**
    *   Updated working draft published. Charter extension in progress.
    *   Moving towards Candidate Recommendation, requiring API stability and two independent implementations (Chrome, Firefox).
    *   Congestion control Constructor argument added (default, throughput, low latency).
    *   Ongoing debate regarding prioritization API complexity.
    *   Stat surface debate focusing on available throughput and time base.

*   **Capsule Design Team:**
    *   Flow control aspects largely removed from pull requests (except for H2 and some H3 text).
    *   Existing TOVs converted to capsules.
    *   Call for thorough review and landing of the design.

*   **Negotiating Web Transport:**
    *   Moving towards independent signaling of support across different layers.
    *   Debate on signaling web transport support via settings. Should the client send a setting, or only the server? Discussing implications of Victor's versioning.

*   **WebTransport over HTTP/3 Open Issues:**
    *   Requires you to opt-in to supporting capsules be quick datagrams C HTTP datagrams and de extended connect.
    *   Discussion around support for HTTP redirects and the challenges of ensuring item potency and handling flow control.
    *   Debate on how to handle bi-directional streams and whether to define bi-directional stream types or use a special frame.

*   **Reliably Resetting Streams:**
    *   Three potential options for handling stream resets were presented:
        *   Doing nothing (reset stream immediately).
        *   Using a reset capsule.
        *   Solving the issue at the QUIC layer.
    *   The preferred approach among participants was to explore solving the issue at the QUIC layer through a new QUIC extension.

## Decisions and Action Items

*   **Capsule Design Team Output:** The output of the design team will be merged.
*   **Negotiating Web Transport:** The discussion about whether clients or servers should signal web transport support via setting will be continued on the mailing list. Specifically, should only the server send `settings web transport Max sessions`?
*   **WebTransport over HTTP/3 Open Issues:** Discussion about how to handle bi-directional streams and whether to define bi-directional stream types or use a special frame will be continued on the mailing list and in a smaller design team.
*   **Reliably Resetting Streams:** The working group will work to address the issue of reliably resetting streams. The current approach is to move the discussion of this to the Quick working group to find a generic QUIC level solution.
*   **Action Item (Bernard):** Track items needed to close out web transport.
*   **Action Item (Chair):** Coordinate with HTTP and QUIC chairs.
*   **Action Item (Lucas):** Consider scheduling a QUIC interim on this topic.

## Next Steps

*   Continue discussions on the mailing lists for open issues.
*   Coordinate with the HTTP and QUIC working groups.
*   Explore potential solutions within the QUIC working group for reliably resetting streams.
*   Address the outstanding PRs.