amethyst/quic/plans/2026-04-26-quic-stack-status.md

QUIC + WebTransport stack — current state (2026-04-26)

This document is the post-implementation snapshot of :quic. It supersedes the original pure-kotlin QUIC + WebTransport plan which was written before any code shipped and is now historical.

TL;DR

:quic is a self-contained Kotlin-Multiplatform module that speaks QUIC v1

• HTTP/3 + WebTransport against real-world servers (aioquic + picoquic verified; nestsClient/MoQ on top). It uses Quartz crypto primitives only — no BouncyCastle, no JNI. ~8.5k lines of production code, ~5k lines of tests, 39 test files, five rounds of parallel audit + fix passes.

What shipped vs the original plan

Phase	Original estimate	Actual	Notes
A. Foundations	1 wk	done	KMP module, UdpSocket on jvmAndroid, Varint migrated from nestsClient
B. TLS 1.3	3 wk	done	RFC 8446 client state machine, X25519 ECDHE, RFC 8448 §3 vectors pass bit-for-bit
C. Initial + Handshake packets	2 wk	done	RFC 9001 §A.2/§A.3 vectors pass; ChaCha20 per §A.5
D. 1-RTT + STREAM	1 wk	done	Stream offset reassembly, FIN, fuzzed
E. ACK + flow control	1 wk	done	MAX_DATA / MAX_STREAM_DATA / MAX_STREAMS routing + writer enforcement
F. Loss recovery + congestion control	1 wk	done (no CC)	RFC 9002 §5/§6 RTT estimator + packet/time-threshold loss detection + PTO + per-frame retransmit shipped 2026-05-05; congestion control still TBD
G. Datagram extension	½ wk	done	RFC 9221 frames + bounded incoming queue
H. Connection lifecycle	1 wk	done	CONNECTION_CLOSE, idle timeout, draining/closing, idempotent driver close
I. HTTP/3	2 wk	done	Control stream + SETTINGS + GOAWAY (with id-regression check) + duplicate-id rejection
J. QPACK	2 wk	done	Static-table + Huffman + integer codec (RFC 7541 §B + RFC 9204 §B.1 vectors)
K. Extended CONNECT + WT	1 wk	done	Stream-type prefixes, HTTP Datagram + WT_CLOSE_SESSION capsule
L. Interop + hardening	2 wk	done + much more	Live interop against aioquic Docker; 5 rounds of audit + fix beyond the plan

The original plan estimated 17–19 weeks. We ran the full sequence plus five unscheduled audit rounds. Every audit found real bugs; the suite is what caught them on regression.

What's actually in the module

quic/
├── plans/                                ← module-local design docs
└── src/
    ├── commonMain/kotlin/com/vitorpamplona/quic/
    │   ├── Buffer.kt                     ← QuicReader / QuicWriter
    │   ├── Varint.kt                     ← RFC 9000 §16
    │   ├── connection/                   ← QuicConnection orchestrator + Driver
    │   │   ├── QuicConnection.kt              (≈ 600 lines, the hub)
    │   │   ├── QuicConnectionDriver.kt        (read/send loops + close)
    │   │   ├── QuicConnectionParser.kt        (feedDatagram + dispatchFrames)
    │   │   ├── QuicConnectionWriter.kt        (drainOutbound + flow-control updates)
    │   │   ├── PacketProtection.kt + builder
    │   │   ├── PacketNumberSpace.kt
    │   │   ├── ConnectionId.kt + TransportParameters.kt
    │   │   └── EncryptionLevel.kt + LevelState.kt
    │   ├── crypto/                       ← Aead, header protection, HKDF helpers, AesEcbHeaderProtection,
    │   │                                   ChaCha20HeaderProtection, ChaCha20Poly1305Aead, InitialSecrets,
    │   │                                   PlatformAesOneBlock, PlatformChaCha20Block (expect),
    │   │                                   bestAes128GcmAead (expect)
    │   ├── frame/                        ← Frame.kt sealed hierarchy + FrameFuzzerTest target
    │   │                                   includes RESET_STREAM / STOP_SENDING / NEW_TOKEN
    │   ├── http3/                        ← Http3FrameReader + Http3Settings + frame types
    │   ├── packet/                       ← LongHeaderPacket, ShortHeaderPacket, RetryPacket, peekHeader
    │   ├── qpack/                        ← QpackDecoder, QpackEncoder, QpackHuffman, QpackInteger,
    │   │                                   QpackStaticTable
    │   ├── connection/recovery/         ← RecoveryToken + SentPacket + QuicLossDetection
    │   │                                   (RFC 9002 §5/§6 RTT estimator + loss detection +
    │   │                                   PTO; per-frame retransmit dispatch)
    │   ├── recovery/                     ← AckTracker (with ack-eliciting gating)
    │   ├── stream/                       ← QuicStream, ReceiveBuffer (with FIN-fully-read), SendBuffer, StreamId
    │   ├── tls/                          ← TlsClient state machine + ClientHello/ServerHello/EE/Cert/CV/Finished
    │   │                                   codecs, TlsKeySchedule, TlsTranscriptHash (incremental), TlsConstants,
    │   │                                   PermissiveCertificateValidator, TlsRunningSha256 (expect)
    │   ├── transport/                    ← UdpSocket (expect)
    │   └── webtransport/                 ← QuicWebTransportFactory + QuicWebTransportSessionState +
    │                                       WtPeerStreamDemux + WtCapsule + WtDatagram + ExtendedConnect
    └── jvmAndroid/kotlin/com/vitorpamplona/quic/
        ├── crypto/JcaAesGcmAead.kt       ← cached JCA Cipher per direction with IV-reuse fallback
        ├── crypto/PlatformCrypto.kt      ← actuals
        ├── tls/JdkCertificateValidator.kt ← system-trust-store chain validation + RSA-PSS / ECDSA / Ed25519
        ├── tls/TlsRunningSha256.kt       ← MessageDigest.clone()-based incremental hash
        └── transport/UdpSocket.kt        ← DatagramChannel + suspend wrapper

Crypto surface

Quartz primitives only:

Primitive	Source
AES-128-GCM	JcaAesGcmAead (jvmAndroid) — cached Cipher per direction; Aes128Gcm singleton (commonMain) for non-hot paths
ChaCha20-Poly1305	Quartz ChaCha20Poly1305 (commonMain pure-Kotlin) wrapped in ChaCha20Poly1305Aead
HKDF-Extract / Expand-Label	Quartz Hkdf + MacInstance; thin RFC 8446 §7.1 helper in crypto/HkdfHelpers.kt
SHA-256 (one-shot)	Quartz sha256(...)
SHA-256 (incremental, for transcript)	TlsRunningSha256 (expect/actual; jvmAndroid wraps MessageDigest.clone())
X25519 ECDHE	Quartz X25519
Ed25519	Quartz Ed25519 (only inside the JVM cert validator path)
AES-ECB (one block, for header protection)	Cipher.getInstance("AES/ECB/NoPadding") (jvmAndroid only)
ChaCha20 keystream (header protection)	Quartz ChaCha20Core.chaCha20Xor
SecureRandom	Quartz RandomInstance

X.509 chain validation, hostname verification, signature verification all delegate to JDK TrustManagerFactory / Signature.getInstance(...). CertificateValidator is a non-null typed parameter — tests pass an explicit PermissiveCertificateValidator; production passes JdkCertificateValidator.

What we deliberately don't do

• QUIC server role. Client-only.
• 0-RTT / session resumption. No PSK extension offered; an arriving ServerFinished without prior Certificate is hard-failed.
• Connection migration / preferred address / multiple paths. NEW_CONNECTION_ID and PATH_* frames decode (so peers don't break us) but aren't acted on.
• Path MTU discovery. Fixed 1200-byte ceiling per RFC 9000 §14.
• HTTP/3 server push.
• QPACK dynamic-table inserts on the encoder. We send literal-only; decoder accepts dynamic-table indexed lines.
• ECN / anti-amplification limits. We're a client.
• TLS Key-Update / NewSessionTicket. Detected and refused (KeyUpdate fails the connection rather than silently desynchronising).
• Congestion control (NewReno / CUBIC / BBR). Loss detection is in place (RFC 9002 §5/§6) but no rate-limiting feedback loop reacts to losses; we send as fast as the application provides bytes. Independent follow-up.

Verified interop

• aioquic (Python): quic-interop-runner-style Docker setup; full handshake + Extended CONNECT + h3 datagram round-trip.
• picoquic (C): Docker image, lightweight HTTP/3 GET.
• In-memory pipe (InMemoryQuicPipe, modeled on Cloudflare quiche's Pipe) drives both sides of the handshake in one JVM for fast tests without sockets.

What's NOT verified: a live nests/MoQ audio-room exchange end to end. That gates on the audio-rooms completion plan (nestsClient/plans/2026-04-26-audio-rooms-completion.md).

Audit summary

Round	Focus	Findings	Status
1	Initial review (pre-interop)	6 critical correctness/security bugs	all fixed
2	TLS hardening + lifecycle	hangs + TLS edge cases	all fixed
3	Performance + concurrency	cipher caching, polling, transcript O(n²)	all fixed
4	Core + TLS + perf + coverage gaps (4 parallel agents)	~30 items including 4 CRITICAL interop blockers	all fixed; comprehensive regression tests added
5	Regression check + concurrency-specific (2 parallel agents)	1 CRITICAL ackEliciting regression I'd just introduced + WT scope leak + others	all fixed

Every fix carries an inline audit-N #M reference comment so the regression test → fix → comment chain is auditable. The whole audit corpus is in the git log (commits whose subject starts with fix(quic): or perf(quic):).

Test inventory

Roughly grouped:

• RFC vectors: RFC 8448 §3 (TLS handshake), RFC 9001 §A.1–A.5 (Initial encrypt/decrypt, Retry, ChaCha20, server-side HP), RFC 9204 §B.1 (QPACK), RFC 7541 (Huffman).
• End-to-end pipe tests: InMemoryQuicPipeTest, CoalescedPacketSkipTest, ReceiveLimitEnforcementTest, PeerStreamLimitTest, FrameRoutingTest, AckElicitingFramesTest.
• Adversarial: FrameFuzzerTest, HostilePacketInputTest, TlsSecurityPropertiesTest, HelloRetryRequestTest.
• Crypto: JcaAesGcmAeadTest, ChaCha20Poly1305AeadTest, TlsTranscriptHashTest.
• WT / HTTP/3: CapsuleReaderTest, WtPeerStreamDemuxTest, WtFramingTest.
• Recovery: AckTrackerCoalescedTest, AckTrackerGatingTest, RecoveryTokenTest, SentPacketTest, ReceiverFlowControlTest (9 cases mirrored from neqo fc.rs), QuicLossDetectionTest, PtoTest, QuicConnectionRetransmitTest, SendBufferRetainUntilAckTest (14 cases for the rewrite), StreamRetransmitTest, CryptoRetransmitTest, ResetStopSendingEmitTest (7 cases).
• Interop: InteropRunner (jvmTest, drives a real socket against a Dockerised aioquic; opt-in, not in CI).

Known limitations / deferred work

These are the items future audit rounds keep flagging that we've consciously not tackled — all confined to the steady-state path that audio rooms don't exercise heavily:

1. No STREAM retransmit on loss (audit-4 #10). Resolved 2026-05-05 — SendBuffer rewritten for retain-until-ACK with three-state range tracking; ACK / loss dispatchers re-queue lost ranges to the retransmit FIFO. Also covers CRYPTO retransmit per encryption level and RESET_STREAM / STOP_SENDING / NEW_CONNECTION_ID retransmit. See 2026-05-04-control-frame-retransmit.md.
2. SendBuffer doesn't retain bytes until ACK. Resolved with #1.
3. No Initial / Handshake key discard. RFC 9000 §17.2.2 / RFC 9001 §4.9 require dropping these after handshake completes; we hold them indefinitely. Memory leak per long session.
4. No path validation for NEW_CONNECTION_ID. We don't migrate.
5. Stateless reset detection. Stateless-reset packets look like corruption to us.
6. AckTracker.purgeBelow threshold semantics. Pre-existing bug: purges based on peer's largestAcknowledged of OUR outbound PNs, but purges OUR inbound PN tracker. Causes range-list bloat, not correctness failure.
7. Driver direct unit tests require turning UdpSocket from expect class into an interface so the test side can stub. The driver is covered indirectly by every pipe-based test plus the live interop runner.
8. No congestion control. Loss detection is wired but nothing throttles send rate in response. Independent ~1-2 wk project.

Pointers

• Original (frozen) plan: docs/plans/2026-04-22-pure-kotlin-quic-webtransport-plan.md
• Audio-rooms NIP draft: docs/plans/2026-04-22-nip-audio-rooms-draft.md
• Completion plan: nestsClient/plans/2026-04-26-audio-rooms-completion.md
• Retransmit plan + implementation log: 2026-05-04-control-frame-retransmit.md
• Live interop runner: quic/src/jvmTest/.../interop/InteropRunner.kt
• Audit history: git log --grep='audit' -- quic/