H3 SETTINGS frame parsing and setting issues

[gmarzot]

it looks like the parsing of the SETTINGS frame needs update. the rfc references appear outdated.

class Setting(IntEnum):
    QPACK_MAX_TABLE_CAPACITY = 0x1
    MAX_FIELD_SECTION_SIZE = 0x6
    QPACK_BLOCKED_STREAMS = 0x7

    # https://datatracker.ietf.org/doc/html/rfc9220#section-5
    ENABLE_CONNECT_PROTOCOL = 0x8
    # https://datatracker.ietf.org/doc/html/rfc9297#section-5.1
    H3_DATAGRAM = 0x33
    # https://datatracker.ietf.org/doc/html/draft-ietf-webtrans-http2-02#section-10.1
    ENABLE_WEBTRANSPORT = 0x2B603742

    # Dummy setting to check it is correctly ignored by the peer.
    # https://datatracker.ietf.org/doc/html/rfc9114#section-7.2.4.1
    DUMMY = 0x21

also there is no way to override issued SETTINGS frame data. this should support override at H3Connection instantiation

In connection with moxygen/proxygen Meta server I see SETTING frame like this

2025-01-31 09:12:34.974 DEBUG: QUIC EVENT: stream 3 data: 0x0004260150000680020000074064080180ffd277014276013301ab6037420140dbc0000000faceb00c
2025-01-31 09:12:34.975 DEBUG: H3 Received Settings:
2025-01-31 09:12:34.975 DEBUG:   Setting 0x1 = 4096
2025-01-31 09:12:34.975 DEBUG:   Setting 0x6 = 131072
2025-01-31 09:12:34.975 DEBUG:   Setting 0x7 = 100
2025-01-31 09:12:34.975 DEBUG:   Setting 0x8 = 1
2025-01-31 09:12:34.975 DEBUG:   Setting 0xffd277 = 1
2025-01-31 09:12:34.975 DEBUG:   Setting 0x276 = 1
2025-01-31 09:12:34.975 DEBUG:   Setting 0x33 = 1
2025-01-31 09:12:34.975 DEBUG:   Setting 0x2b603742 = 1
2025-01-31 09:12:34.975 DEBUG:   Setting 0xdb = 4207849484

current proxygen settings

[gmarzot]

I also observe that version compliance with various draft-ietf-webtrans-http3 draft versions is confusing . For instance the ENABLE_WEBTRANSPORT setting is no longer defined in the latest version of the draft. I can not tell what the "correct" set is .

So perhaps this issue is only a feature request to allow initial settings to be provided to H3Connection to support custom SETTINGS contents.

[jlaine]

Yeah this is annoying. I was hoping by now there would be an actual published RFC instead of this endless stream of drafts. I'll see if I can find time to read the updated drafts.