Enum DraftVersion

Source

pub enum DraftVersion {
Show 15 variants    Draft07,
    Draft08,
    Draft09,
    Draft10,
    Draft11,
    Draft12,
    Draft13,
    Draft14,
    Draft15,
    Draft16,
    Draft17,
    Draft18,
    Draft19,
    Draft20,
    Draft21,
}

Expand description

MoQT draft version for runtime codec selection.

Variants§

§

Draft07

draft-ietf-moq-transport-07.

§

Draft08

draft-ietf-moq-transport-08.

§

Draft09

draft-ietf-moq-transport-09.

§

Draft10

draft-ietf-moq-transport-10.

§

Draft11

draft-ietf-moq-transport-11.

§

Draft12

draft-ietf-moq-transport-12.

§

Draft13

draft-ietf-moq-transport-13.

§

Draft14

draft-ietf-moq-transport-14.

§

Draft15

draft-ietf-moq-transport-15.

§

Draft16

draft-ietf-moq-transport-16.

§

Draft17

draft-ietf-moq-transport-17.

§

Draft18

draft-ietf-moq-transport-18.

§

Draft19

draft-ietf-moq-transport-19.

§

Draft20

draft-ietf-moq-transport-20.

§

Draft21

draft-ietf-moq-transport-21.

Implementations§

Source §

impl DraftVersion

Source

pub const ALL: [DraftVersion; 15]

Every draft of the series, oldest first.

The variants are not feature-gated, so this is the whole series whatever the build compiles. It answers which drafts exist, which is a property of the specification; which drafts this binary can decode is a different question with a different answer per feature set, and moqtap-proxy’s draft_is_compiled is where that one lives.

Sweeps should read this rather than write their own list. A written-out draft list is the most expensive silent defect this workspace has: it compiles, it passes, and it tests one draft fewer than it claims to. shape/matcher.rs shipped a [DraftVersion; 13] under a doc comment claiming the whole series, and every unit test that swept it stopped covering the newest draft without failing. A sweep over ALL cannot do that, and a sweep that deliberately covers less than the series – one draft per era, or the drafts that have a joining fetch - still writes its own list and still says why.

An array rather than a slice, so for d in DraftVersion::ALL yields drafts and not references and a caller can name the type. The length beside it is checked by the compiler against the contents, so it cannot silently disagree with them - and the contents are the half that goes wrong. scripts/check-draft-parity.py holds those against the enum, the per-draft source directories, the cargo features and the CI rows on every run, and the tests below hold them against from_number.

Source

pub const fn newest() -> DraftVersion

The newest draft of the series.

ALL is ordered, so this is its last element. Written as a method rather than left to the caller because ALL.last().unwrap() in a hundred places is a hundred unwraps, and because a caller that wants “the newest” almost always wants it infallibly.

Source

pub fn version_varint(&self) -> VarInt

The MoQT version number this draft would announce in CLIENT_SETUP.

Format: 0xff000000 + draft_number.

From draft-15 on there is no such value on the wire at all. Draft-15 deleted the version field from CLIENT_SETUP and moved version selection into the ALPN (moqt-<N>, see Self::quic_alpn), so the number this returns for drafts 15 through 21 — 0xff00000f through 0xff000015 — is a continuation of the mapping and not something a peer can observe or send. Nothing in this crate encodes it for those drafts. It is kept so that a caller with a draft in hand can name the version the series would have used, and so the mapping does not acquire a hole.

A tool that tries to detect the negotiated draft by looking for 0xff0000NN in a capture will find nothing from draft-15 on; the ALPN is the only signal.

Source

pub fn quic_alpn(&self) -> &'static [u8] ⓘ

The ALPN protocol identifier for raw QUIC connections.

Drafts 07–14 all use moq-00 and negotiate the draft version in CLIENT_SETUP / SERVER_SETUP. Draft-15+ encode the draft number in the ALPN itself (moqt-<N>), so version selection happens during the TLS handshake rather than after it.

Source

pub fn from_alpn(alpn: &[u8]) -> Option<DraftVersion>

Resolve an ALPN identifier to a specific draft version.

Returns Some for ALPNs that unambiguously identify a draft (moqt-15 through moqt-21). Returns None for moq-00 — which covers drafts 07–14 and requires inspecting CLIENT_SETUP’s supported-versions list — and for any unrecognized ALPN.

Source

pub fn from_number(n: u8) -> Option<DraftVersion>

Resolve a draft number (e.g. 7..=21) to a DraftVersion.

Returns None for numbers outside the supported range.

Source

pub fn uses_fixed_length_framing(&self) -> bool

Whether this draft uses a 16-bit big-endian message length in control message framing (true) or a QUIC varint (false).

Draft-11 changed the framing from Length(i) to Length(16).

Source

pub fn varint_encoding(&self) -> VarIntEncoding

Which variable-length integer encoding this draft’s wire format uses.

Matched draft by draft rather than derived from the number. The series has already changed encoding once mid-stream and revised it again a draft later, so there is no rule to extrapolate from: adding a variant to DraftVersion must fail to compile here until someone reads that draft and says which encoding it uses.

Source

pub fn uses_moqt_varint(&self) -> bool

Whether this draft uses one of MoQT’s own variable-length integers rather than RFC 9000’s.

Source

pub fn varint_len(&self, first_byte: u8) -> usize

The total encoded length of a variable-length integer, from its first byte, under this draft’s encoding.

Available without a buffer, because a reader needs it to know how many bytes to wait for before it can decode at all. On draft-17 a first byte of 11111100 reports 7 even though the draft forbids that length: the reader waits for the whole field, then Self::decode_varint rejects it.

Source