--- status: reviewed last_updated: 2026-07-15 --- # alknet-tls Shared TLS configuration and certificate management — server and client. Builds a `rustls::ServerConfig` (or an ACME state machine + cert resolver) once and shares it across multiple transports — quinn, `tokio-rustls` (TCP+TLS), and iroh — so one certificate identity serves QUIC and TCP endpoints simultaneously. Builds a `rustls::ClientConfig` with ADR-034 verifier selection and ADR-084 crypto provider, shared across all outbound-dialing crates (hub, worker, `CallClient`, `ChannelClient`). One ACME state machine, one cert, N transports; one verifier rule, N clients. ## What `alknet-tls` extracts the TLS setup that was welded to the quinn endpoint in `alknet-core`. The existing code (`endpoint.rs`) builds a `rustls::ServerConfig` from a `TlsIdentity`, then **consumes** it into a `quinn::ServerConfig` — making it impossible to reuse the same cert for a TCP+TLS listener. ACME is worse: the `AcmeState` task is spawned inside the quinn endpoint, so a TCP+TLS listener would need its own ACME state machine (two orders for the same domain, two cert caches, potential Let's Encrypt rate-limiting). `alknet-tls` fixes this by making the TLS config **shareable**: ```rust pub struct TlsServerConfig { config: rustls::ServerConfig, // Clone-safe — Arc internally acme_handle: Option>, // one ACME task (see lifecycle below) } impl TlsServerConfig { pub async fn new(identity: &TlsIdentity, alpns: &[Vec]) -> Result; /// Produce a quinn server config. Clones the inner rustls config /// (cheap — Arc-shared cert resolver) and wraps it for quinn. /// Feature-gated on `quinn`. #[cfg(feature = "quinn")] pub fn for_quinn(&self) -> Result; /// Produce a tokio-rustls acceptor for TCP+TLS. Clones the inner /// rustls config. Feature-gated on `tcp`. #[cfg(feature = "tcp")] pub fn for_tcp_tls(&self) -> tokio_rustls::TlsAcceptor; /// Borrow the underlying rustls config, for any other consumer. pub fn rustls_config(&self) -> &rustls::ServerConfig; } ``` `TlsServerConfig` is **not `Clone`** — it holds a `JoinHandle` for the ACME task, which is not cloneable. Share it via `Arc`. Each accessor (`for_quinn`, `for_tcp_tls`) clones the inner `rustls::ServerConfig`, which is cheap (it holds `Arc`s to the cert resolver and verifier, not the raw key material). The assembly layer builds one `TlsServerConfig`, wraps it in `Arc`, and hands `Arc::clone()` to each transport consumer. One cert, one ACME state machine, N transports. ## Why `alknet-core` builds the `rustls::ServerConfig` once, then consumes it into a `quinn::ServerConfig` — making the cert unreusable for a TCP+TLS listener. For ACME the problem is worse: the `AcmeState` task is spawned inside the quinn endpoint, so a TCP+TLS listener would need a second ACME state machine for the same domain (duplicate orders, divergent cert caches, Let's Encrypt rate-limit risk). The full rationale, including the cert-reuse problem, the ACME worst case, and the three reasons a separate crate is the right shape (dependency isolation, ACME weight, quinn/iroh having their own TLS), is in [ADR-082](../../decisions/082-alknet-tls-extraction.md). ### The three endpoint types (ADR-086) A hub composes a subset of three endpoint types, each with its own identity model and transport(s). `alknet-tls` provides the `TlsServerConfig`s; the assembly layer builds one per endpoint type that uses a `rustls::ServerConfig` (iroh is the exception — it has its own TLS). | Endpoint type | Identity | `TlsServerConfig` | Transport(s) | Browsers? | |---------------|----------|-------------------|--------------|-----------| | **native** | RFC 7250 raw key (Ed25519) | raw-key config | QUIC (primary), TCP+TLS (fallback when UDP blocked) | No (browsers can't do raw keys) | | **web** | X.509 (manual or ACME) | X.509/ACME config | TCP+TLS (HTTP, WebSocket), QUIC (WebTransport — deferred) | Yes (via HTTPS / WebSocket; WebTransport when revived) | | **iroh** | RFC 7250 raw key (NodeId) | (no `TlsServerConfig` — iroh has its own TLS) | iroh (relay-assisted QUIC) | No | | Development | Self-signed | self-signed config | Any | No (untrusted) | The ALPN list each `TlsServerConfig` advertises is **split by endpoint type** (ADR-086 §3, resolving OQ-62): the native config advertises the native ALPNs (`alknet/channels`, `alknet/call`, `alknet/ssh` future); the web config advertises the entry-point ALPNs (`h2`, `http/1.1`) + `alknet/channels` (for WebSocket-carrying-channels, OQ-65) + `acme-tls/1` (appended automatically). The assembly layer filters `registry.alpn_strings()` per config. See [ADR-086](../../decisions/086-endpoint-types-and-entry-points.md) for the full ALPN-list table and the entry-point/endpoint distinction. In all cases, TLS + ALPNs "just works" — the TLS handshake negotiates the ALPN, the `HandlerRegistry` dispatches by ALPN, the transport is a parameter. The TLS crate's job is to make the cert available to whichever transports the deployment runs. ## Architecture ### What moves from `alknet-core` to `alknet-tls` (server side) | Component | Current location | New location | |-----------|-----------------|-------------| | `TlsIdentity` enum | `alknet-core/config.rs` | **stays in core** (it's a config type) | | `Ed25519SecretKey` | `alknet-core/config.rs` | **stays in core** (config type) | | `build_rustls_server_config()` | `alknet-core/endpoint.rs` (`#[cfg(feature = "quinn")]`) | `alknet-tls` (unconditional) | | `build_quinn_server_config_from_rustls()` | `alknet-core/endpoint.rs` (`#[cfg(feature = "quinn")]`) | `alknet-tls` (`for_quinn()` — wraps rustls config in `QuicServerConfig`) | | `TlsSetup` (ACME state machine) | `alknet-core/endpoint.rs` (`#[cfg(feature = "quinn")]`) | `alknet-tls` (the `TlsServerConfig::new` ACME path) | | `RawKeyCertResolver` | `alknet-core/endpoint.rs` (`#[cfg(feature = "quinn")]`) | `alknet-tls` | | `Ed25519SigningKey` | `alknet-core/endpoint.rs` (`#[cfg(feature = "quinn")]`) | `alknet-tls` (consolidates with the `alknet-call` duplicate — see client table) | | `AcceptAnyCertVerifier` | `alknet-core/endpoint.rs` (`#[cfg(feature = "quinn")]`) | `alknet-tls` | | `SelfSignedCert` / `generate_self_signed_cert()` | `alknet-core/endpoint.rs` (`#[cfg(feature = "quinn")]`) | `alknet-tls` | | `load_cert_chain()` / `load_private_key()` | `alknet-core/endpoint.rs` | `alknet-tls` (consolidates with the `alknet-call` duplicate — see client table) | | `fingerprint.rs` | `alknet-core/fingerprint.rs` | **stays in core** (shared by server + client; the client-side `FingerprintPinVerifier` is now in `alknet-tls` per ADR-089 §5, so both consumers are co-located; production code uses `sha2` + manual DER only — `rustls` is test-only. See OQ-59 — the original dep-edge concern that motivated keeping `fingerprint.rs` in core is dissolved by ADR-089 §5.) | ### What moves from `alknet-call` to `alknet-tls` (client side) `TlsClientConfig::new` (ADR-087) centralizes the client-side verifier selection + provider wiring + client-auth cert presentation that currently lives in `alknet-call/src/client/call_client.rs`. The extraction is the client-side analogue of the server-side `endpoint.rs` extraction above. | Component | Current location | New location | |-----------|-----------------|-------------| | `build_quinn_client_config()` | `alknet-call/client/call_client.rs` (`#[cfg(feature = "quinn")]`) | `alknet-tls` (`TlsClientConfig::new` + `for_quinn()`) | | `build_client_auth()` | `alknet-call/client/call_client.rs` | `alknet-tls` (client-auth cert resolver construction inside `TlsClientConfig::new`) | | `select_server_verifier()` | `alknet-call/client/call_client.rs` | `alknet-tls` (ADR-034 verifier selection inside `TlsClientConfig::new`) | | `load_platform_root_cert_store()` | `alknet-call/client/call_client.rs` | `alknet-tls` (the unknown-X.509-remote CA path inside `TlsClientConfig::new`) | | `FingerprintPinVerifier` | `alknet-call/client/call_client.rs` | `alknet-tls` (moved — it is a TLS concern; `TlsClientConfig::new` constructs it; moving it lets `alknet-call` shed its direct `rustls` dep entirely per ADR-089 §5) | | `Ed25519SigningKey` (client-side copy) | `alknet-call/client/call_client.rs` | `alknet-tls` (consolidates with the `endpoint.rs` duplicate — one copy in `alknet-tls`) | | `RawKeyClientCertResolver` | `alknet-call/client/call_client.rs` | `alknet-tls` | | `NoClientCertResolver` | `alknet-call/client/call_client.rs` | `alknet-tls` | | `load_cert_chain()` / `load_private_key()` (client-side copies) | `alknet-call/client/call_client.rs` | `alknet-tls` (consolidates with the `endpoint.rs` duplicate — one copy in `alknet-tls`) | | `CallClient::connect` | `alknet-call/client/call_client.rs` | **removed** (ADR-089 §5 — the dial is extracted to `AlknetClient`; `CallClient` keeps only `spawn_dispatch`, shedding its TLS/transport deps) | **Consolidation note.** `Ed25519SigningKey` and `load_cert_chain`/`load_private_key` are currently **duplicated** across `endpoint.rs` (server) and `call_client.rs` (client). After extraction there is one copy of each in `alknet-tls`, used by both `TlsServerConfig::new` and `TlsClientConfig::new`. Both call sites (`endpoint.rs`'s server path, `call_client.rs`'s client path) are updated to import from `alknet-tls`. `TlsIdentity` and `Ed25519SecretKey` stay in core because they're config types — `StaticConfig` holds a `TlsIdentity`, and config types belong in core. `alknet-tls` re-exports them for convenience. `fingerprint.rs` stays in core because it's shared by both the server path (endpoint extracts fingerprint from the client cert) and the client path (`FingerprintPinVerifier` — now in `alknet-tls` per ADR-089 §5 — matches the server's cert against a pinned fingerprint). The production code in `fingerprint.rs` uses only `sha2` and manual DER parsing — the `rustls::sign` usage is in the test helper only. See OQ-59 (the original dep-edge concern that motivated keeping `fingerprint.rs` in core is dissolved by ADR-089 §5 — `FingerprintPinVerifier` moved to `alknet-tls`, so its consumers are co-located). ### `TlsServerConfig` The central type. Built once from a `TlsIdentity` + ALPN list, shared across transports. ```rust pub struct TlsServerConfig { /// The rustls server config. Clone-safe (holds Arcs to cert resolver /// and verifier, not raw key material). config: rustls::ServerConfig, /// The ACME state machine task, if ACME is active. One task, shared /// — dropping this handle does NOT stop the ACME state machine /// (it's owned by the config, not the endpoint). acme_handle: Option>, } ``` Construction: ```rust impl TlsServerConfig { /// Build a TLS server config from the given identity and ALPN list. /// For ACME, spawns the ACME state machine task and wires its /// resolver into the rustls config. For X509/RawKey/SelfSigned, /// loads the cert and builds the resolver directly. pub async fn new( identity: &TlsIdentity, alpns: &[Vec], ) -> Result; } ``` The ALPN list is the set of ALPNs the endpoint type advertises. For a native config: `alknet/channels`, `alknet/call`, `alknet/ssh` (future). For a web config: `h2`, `http/1.1`, `alknet/channels` (for WebSocket-carrying-channels, OQ-65). The list is **split by endpoint type** (ADR-086 §3) — the assembly layer filters `registry.alpn_strings()` per `TlsServerConfig`, not passes the same list to both. For ACME, the `acme-tls/1` ALPN is appended automatically (for the TLS-ALPN-01 challenge, ADR-027 §7). ### `async fn new` — lifecycle semantics `new` is `async` because the ACME path spawns a state-machine task (`tokio::spawn`) before returning — the spawn itself is await-free, but the function is async so the non-ACME paths share one signature. **ACME path**: `new` spawns the `AcmeState` task, wires its `resolver()` into the `rustls::ServerConfig`, and returns **immediately** — it does **not** await the first certificate. The returned `TlsServerConfig` is usable for `for_quinn()` / `for_tcp_tls()` right away; the resolver may return no cert until the first ACME order completes, causing TLS handshakes to fail transiently during that window. This matches the current code's behavior (`TlsSetup::new_acme` spawns and returns). **Non-ACME paths** (X509 / RawKey / SelfSigned): cert loading is synchronous file I/O (`std::fs::read`) + in-memory construction; there is no await point in the implementation. The `async` signature is for API uniformity with the ACME path, not because the work is async. An implementer who finds this objectionable may split a non-async constructor — that is a two-way-door implementation detail, not an architecture decision. ### Behavior-preservation invariants The extraction must preserve these load-bearing TLS behaviors. They originate from [ADR-027](../../decisions/027-tls-identity-redesign-acme-rawkey-decoupling.md), which established the `TlsIdentity` model, the `Acme` variant, and the `acme-tls/1` ALPN challenge handling. An implementer who omits any of these produces a crate that compiles and passes type-checks but silently changes TLS behavior: - **`max_early_data_size = u32::MAX`** on all server config paths (X509, RawKey, SelfSigned, ACME). Enables 0-RTT / early data. Omitting it disables 0-RTT, silently breaking clients that use it. - **`rustls::crypto::aws_lc_rs::default_provider()`** as the crypto provider on all paths. Do not switch to `ring` or the process-default provider without a new ADR — see [ADR-084](../../decisions/084-aws-lc-rs-crypto-provider.md) for the rationale (FIPS, platform matrix, iroh consistency). - **`AcceptAnyCertVerifier`'s `supported_verify_schemes()`** returns ED25519 + ECDSA P-256/P-384 + RSA PSS/PKCS1 (SHA256/384/512). This list determines which client cert signature algorithms the server accepts. Must be preserved verbatim. - **`acme-tls/1` ALPN append** for the ACME path only (ADR-027 §7). The TLS-ALPN-01 challenge requires the server to advertise `acme-tls/1` in its ALPN list. Appended in `TlsServerConfig::new`'s ACME branch, not by the caller. Transport-specific accessors: ```rust impl TlsServerConfig { /// Produce a `quinn::ServerConfig` for a QUIC listener. Clones the /// rustls config (cheap — Arc-shared cert resolver), wraps it in /// `QuicServerConfig`. Returns `Result` because /// `QuicServerConfig::try_from(rustls::ServerConfig)` can fail if /// the rustls config contains quinn-incompatible settings. /// Feature-gated on `quinn`. #[cfg(feature = "quinn")] pub fn for_quinn(&self) -> Result; /// Produce a `tokio_rustls::TlsAcceptor` for a TCP+TLS listener. /// Clones the rustls config. Infallible — /// `TlsAcceptor::new(rustls::ServerConfig)` cannot fail. /// Feature-gated on `tcp` (pulls `tokio-rustls`). #[cfg(feature = "tcp")] pub fn for_tcp_tls(&self) -> tokio_rustls::TlsAcceptor; /// Borrow the underlying rustls config, for consumers that need to /// build their own transport-specific wrapper not covered by /// `for_quinn` / `for_tcp_tls`. No current consumer (iroh reads the /// `Ed25519SecretKey` directly, not the rustls config — see "Iroh: /// shares the key, not the rustls config" below); kept as a /// forward-looking accessor for future transport wrappers. pub fn rustls_config(&self) -> &rustls::ServerConfig; } ``` ### Iroh: shares the key, not the rustls config Iroh is different from quinn and TCP+TLS: it has its own TLS built into the `Endpoint`, using RFC 7250 raw keys. It does not consume a `rustls::ServerConfig` — it takes an `iroh::SecretKey` and handles TLS internally. So `alknet-tls` does not have a `for_iroh()` method. Instead, the assembly layer reads the `Ed25519SecretKey` from `StaticConfig` (stays in core) and passes it to iroh's `Endpoint::builder().secret_key()` directly. `alknet-tls` is involved only when iroh is not the sole transport — in that case, the same `Ed25519SecretKey` feeds both `TlsServerConfig::new(TlsIdentity::RawKey(key), ...)` (for quinn/TCP) and `iroh::SecretKey::from_bytes(key.as_bytes())` (for iroh). The fingerprint is normalized across all three paths (ADR-030 §6): `ed25519:` for raw keys, whether the cert came from quinn's `RawKeyCertResolver`, iroh's built-in TLS, or a future TCP+TLS raw-key path. `fingerprint.rs` (in core) handles this. ### Feature gates ```toml [features] default = [] quinn = ["dep:quinn"] # for_quinn() — wraps rustls config for quinn tcp = ["dep:tokio-rustls"] # for_tcp_tls() — wraps rustls config for TCP+TLS acme = ["dep:rustls-acme"] # ACME state machine ``` A deployment that only uses quinn enables `quinn`. A deployment that uses TCP+TLS enables `tcp`. A deployment that uses both enables both. ACME is opt-in (heavy dep, long-running task). The `rustls` dep is always present (it's the core TLS library). ### Dependencies ``` alknet-tls ├── alknet-core (TlsIdentity, Ed25519SecretKey, fingerprint — re-exported) ├── rustls (ServerConfig, ClientConfig, cert types — always present) ├── rustls-pki-types (CertificateDer, PrivateKeyDer, etc. — via rustls re-export │ or direct dep; core lists it directly) ├── rustls-pemfile (cert/key file loading — always present) ├── rustls-native-certs (platform root cert store — always present; the │ unknown-X.509-remote CA path in `TlsClientConfig::new`) ├── webpki-roots (built-in CA roots fallback — always present; merged │ into the root store when the platform store is empty, │ so a containerized deployment with no system CA bundle │ can still verify public X.509 remotes — see ADR-088 §5) ├── rcgen (self-signed cert generation — always present) ├── ed25519-dalek (Ed25519 signing key — always present, via core) ├── sha2 (fingerprint computation — always present, via core) ├── tokio (spawn for ACME task — always present) ├── futures (StreamExt for ACME event loop — acme-gated) ├── tracing (logging) ├── quinn (optional — for_quinn()) ├── tokio-rustls (optional — for_tcp_tls()) └── rustls-acme (optional — ACME state machine) ``` `rustls-native-certs` and `webpki-roots` are always-present deps (not feature-gated) because the unknown-X.509-remote CA-verification path in `TlsClientConfig::new` is needed by any client dialing a public X.509 endpoint, regardless of transport (QUIC or TCP+TLS). In the pre-extraction code these lived in `alknet-call` behind the `quinn` feature; the extraction (ADR-089 §5) moves them to `alknet-tls` ungated, and `alknet-call` sheds the deps entirely. `alknet-core` loses `rustls-pemfile`, `rcgen`, and `rustls-acme` from its dependencies — the cert-loading, self-signed generation, and ACME machinery move to `alknet-tls`. Core's `acme` feature (`acme = ["dep:rustls-acme"]` in `Cargo.toml` and the `#[cfg(feature = "acme")]` gates on `acme_state_handle` in `endpoint.rs`) becomes vestigial after the extraction and is removed — the ACME state machine now lives on `TlsServerConfig` in `alknet-tls`, not on `AlknetEndpoint`. Core keeps `quinn` and `iroh` (the endpoint struct and accept loops remain in core), `ed25519-dalek` (`Ed25519SecretKey` stays in `config.rs`), and `rustls` / `rustls-pki-types` (`fingerprint.rs` uses `rustls::pki_types` in production and `rustls::sign` in the test helper `build_ed25519_spki_der` — see OQ-59). > **Terminology — hub, worker, hub-worker.** A *hub* is a node that > accepts inbound connections from workers and browsers (the central > node in a hub-and-spoke topology — see > [`crates/hub/README.md`](../hub/README.md)). A *worker* is a node > that dials out to a hub. A *hub-worker* is a node that does both > (accepts inbound and dials out). A *pure worker* has no inbound > endpoints. These terms come from the hub topology (ADR-029, ADR-034); > "assembly layer" (ADR-014) is the deployment binary that wires crates > — in practice, today, usually a hub or hub-worker. ### Implementation ordering `alknet-tls` is greenfield — `crates/alknet-tls` does not exist yet. The endpoint section below ("What `AlknetEndpoint` does after the refactor") describes the **post-refactor target**, not the current source. The current `crates/alknet-core/src/endpoint.rs` is the **extraction source** — `AlknetEndpoint::new(static_config, ...)` builds TLS internally, the shape ADR-083 replaces. The endpoint is extracted into a new crate `alknet-endpoint` (ADR-083 Amendment 2026-07-15) as part of this work. The extraction and refactor are **sequenced**, not simultaneous: 1. **`alknet-tls` first** — build the crate in isolation. `TlsServerConfig` and `TlsClientConfig` are unit-testable against `TlsIdentity` without touching the endpoint. This is the greenfield step. 2. **`alknet-endpoint` second** — build the new endpoint crate fresh against the ADR-083 shape (`new(handlers, dynamic, identity_provider, drain_timeout)` + `with_quinn` / `with_iroh` / `with_tcp_tls`), importing `Connection`/`ProtocolHandler`/`AuthContext` from `alknet-core` and taking pre-built transports (no TLS config — the assembly layer builds those via `alknet-tls`). The old `crates/alknet-core/src/endpoint.rs` is deleted. 3. **Assembly layer last** — the deployment binary (hub/worker) builds the `TlsServerConfig`(s) and `TlsClientConfig`(s), the transports, and hands them to `AlknetEndpoint` (in `alknet-endpoint`) via the builder methods. A compilable intermediate state exists after step 1: `alknet-tls` built and tested standalone, with `endpoint.rs` still in its old shape. The call sites for `TlsServerConfig` / `TlsClientConfig` do not exist until step 2/3 — an implementer testing step 1 writes tests against the TLS types directly, not against a wired-up endpoint. ### What `AlknetEndpoint` (in `alknet-endpoint`) does after the refactor `AlknetEndpoint::new()` currently builds `TlsSetup` internally. After the refactor (see [ADR-083](../../decisions/083-endpoint-as-accept-loop-runner.md)), the endpoint (extracted into `alknet-endpoint` per ADR-083 Amendment 2026-07-15) takes **no TLS config at all** — it is a multi-transport accept-loop runner. TCP+TLS is an owned transport (via `with_tcp_tls`), not an external loop: ```rust impl AlknetEndpoint { pub fn new( handlers: HandlerRegistry, dynamic: Arc>, identity_provider: Arc, drain_timeout: Duration, ) -> Self; pub fn with_quinn(mut self, endpoint: quinn::Endpoint) -> Self; pub fn with_iroh(mut self, endpoint: iroh::Endpoint) -> Self; /// TCP+TLS is a first-class owned transport — same `run()` loop, /// same `shutdown()` as quinn/iroh. Feature-gated on `tcp`. #[cfg(feature = "tcp")] pub fn with_tcp_tls( mut self, listener: tokio::net::TcpListener, acceptor: tokio_rustls::TlsAcceptor, ) -> Self; /// Public for SSH channels / future WT (connection-internal /// multiplexing, not listener transports). pub fn dispatch( &self, connection: Connection, alpn: Vec, fingerprint: Option, remote_addr: Option, ); pub async fn run(self: Arc); pub async fn shutdown(&self); } ``` The assembly layer builds the `TlsServerConfig`(s), builds the transports (`for_quinn()` → `quinn::Endpoint::server()`, `for_tcp_tls()` → `TlsAcceptor` paired with a `TcpListener`, `Ed25519SecretKey` → iroh), and hands them to `AlknetEndpoint` via builder methods. A hub serving native clients and browsers holds two `TlsServerConfig`s (raw key + X.509/ACME); the endpoint takes neither — it takes the already-built transport endpoints. The TCP+TLS listener is owned by the endpoint via `with_tcp_tls`; the endpoint runs its accept loop inside `run()` and stops it on `shutdown()`. The ACME handle lives on the `TlsServerConfig`, not the endpoint. This resolves the single-`Arc` problem: the endpoint has no "the TLS config" to take because a hub has two. It also means shutdown is single-owner — the endpoint owns all its accept loops (quinn, iroh, TCP+TLS); one `shutdown()` stops them all. ### The TCP+TLS accept loop (out of scope for this crate) `alknet-tls` provides `for_tcp_tls() -> TlsAcceptor`. The actual TCP accept loop (`TcpListener::accept` → `TlsAcceptor::accept` → `Connection::from_bidi` → `endpoint.dispatch()`) lives in `alknet-endpoint` behind a `tcp` feature, as an owned transport on `AlknetEndpoint` (via `with_tcp_tls(listener, acceptor)` — see ADR-083, Amendment 2026-07-15). `alknet-tls` is the cert provider, not the accept loop. This keeps `alknet-tls` focused on TLS setup and cert sharing, not transport accept logic. ### Client-side — `TlsClientConfig` (ADR-087) `alknet-tls` provides a client-side config alongside `TlsServerConfig`. A hub dials out to workers it supervises and to other hubs (hub-as-client); `alknet-worker` dials a hub. Both need a `rustls::ClientConfig` with ADR-034's verifier selection and ADR-084's crypto provider. `TlsClientConfig` centralizes this — it is a present prerequisite for the first hub deployment, consumed by `AlknetClient`'s QUIC and TCP+TLS dials (ADR-089). There are exactly two clients in the alknet client surface as far as `TlsClientConfig` and `AlknetClient` are concerned — **call** (`CallClient`) and **channels** (`ChannelClient`, which is a proxy over many ALPNs via channel 0). Both must support all three transport accessors below; the TLS config is shared across them, the dial is per-transport per-client. ```rust pub struct TlsClientConfig { config: rustls::ClientConfig, } impl TlsClientConfig { /// Build a client TLS config. Takes two inputs, both derived from /// `Capabilities` (ADR-014) / `ConnectionCredentials`-shaped values /// (ADR-091): /// /// 1. `local_identity` — the local node's `TlsIdentity` (RFC 7250 /// raw key or X.509), presented as the client cert. `None` → /// no client cert (the server gets nothing to fingerprint). /// `SelfSigned` → no client cert (dev-only). `Acme` → /// `TlsError::AcmeConfig` (server-only identity). /// /// 2. `verifier_context` — the inputs to ADR-034's server-cert /// verifier selection: /// - known peer (PeerEntry present) → fingerprint pin /// (FingerprintPinVerifier) /// - unknown remote + X.509 → CA verification /// (WebPkiServerVerifier) /// - unknown remote + raw key → fail closed at handshake (not /// a `new`-time error; see ADR-088 §6) /// /// Applies ADR-084 crypto provider (aws_lc_rs::default_provider()). pub fn new( local_identity: &Option, verifier_context: &ClientVerifierContext, ) -> Result; /// Produce a `quinn::ClientConfig` for a QUIC dial. Clones the /// inner rustls config, wraps it in `QuicClientConfig`. Returns /// `Result` because `QuicClientConfig::try_from(rustls::ClientConfig)` /// can fail with `NoInitialCipherSuite` — the same failure the /// server-side `for_quinn()` surfaces as `TlsError::QuinnWrap`. /// Feature-gated on `quinn`. #[cfg(feature = "quinn")] pub fn for_quinn(&self) -> Result; /// Produce a `tokio_rustls::TlsConnector` for a TCP+TLS dial. /// Clones the inner rustls config. Infallible — /// `TlsConnector::new(rustls::ClientConfig)` cannot fail. /// Feature-gated on `tcp` (pulls `tokio-rustls`). #[cfg(feature = "tcp")] pub fn for_tcp_tls(&self) -> tokio_rustls::TlsConnector; /// Borrow the underlying rustls config, for consumers that need to /// build their own transport-specific wrapper not covered by /// `for_quinn` / `for_tcp_tls` (e.g. a future transport). Not /// feature-gated — returns the raw rustls config, not a /// transport-specific wrapper. pub fn rustls_config(&self) -> &rustls::ClientConfig; } ``` The `ClientVerifierContext` carries the inputs to ADR-034's verifier selection (whether a `PeerEntry` exists for the remote, the expected fingerprint). The exact struct shape is an implementation detail; the decisions are in ADR-034. `ClientVerifierContext` is derived from `ConnectionCredentials` (in `alknet-core`, per ADR-091) at the dial site — `AlknetClient` extracts the TLS-relevant fields (`local_identity` → client cert, `remote_identity` → fingerprint-pin input) and builds a `ClientVerifierContext` from the latter. The call-protocol `auth_token` is not in `ConnectionCredentials` — it is a per-request field on `call.requested` payloads (a call-protocol / hub concept), not a transport credential; it never reaches `TlsClientConfig` or `ClientVerifierContext`. The `TlsError` variant granularity (covering both server and client errors) is decided — see [ADR-088](../../decisions/088-tlserror-shape.md) and the [`TlsError`](#tlserror) section below. **Root store fallback (ADR-088 §5).** The unknown-X.509-remote CA-verification path loads the platform's native root certs (`rustls-native-certs`). If the platform store is empty (e.g. a containerized deployment with no system CA bundle), the built-in `webpki-roots` are merged in so the store is never empty. This makes the `NoRootAnchors` failure mode unreachable in practice — a containerized worker dialing a public X.509 hub succeeds without requiring the operator to mount a CA bundle. Native-certs *load* errors are logged, not returned (preserved behavior); the fallback guarantees the store is non-empty regardless. See ADR-088 §5. `TlsClientConfig` produces a `rustls::ClientConfig`; the caller (the transport-specific dial helper — `AlknetClient::dial_quic` / `dial_tcp_tls`, ADR-089) passes it to the transport's connector. The config is transport-agnostic; the dial is not. This is the client-side analogue of ADR-065's server-side separation: the take-over (`spawn_dispatch` / `from_connection`, transport-agnostic) is built now; the dial (transport-specific) is per-transport. The transport-polymorphic dial is now extracted as `alknet-client` (ADR-089, resolves OQ-55) — `AlknetClient` builds the `TlsClientConfig` per-dial and calls the transport's connector. The client-side accessor API mirrors the server side: `for_quinn()` / `for_tcp_tls()` / `rustls_config()` — three transports, same pattern. Iroh is the exception (see below). `AlknetClient` (ADR-089) consumes `TlsClientConfig` via these accessors for the QUIC and TCP+TLS dials; the iroh dial is the key-not-config exception. ### Iroh — shares the key, not the config (client side too) Iroh's client side, like its server side (above), does not consume a `rustls::ClientConfig` — it takes an `iroh::SecretKey` and handles TLS internally. The iroh client dial does not use `TlsClientConfig`. The verifier selection for iroh is fingerprint-pinning by another name: iroh's built-in TLS verifies the remote's `NodeId` (Ed25519 public key) against the expected `NodeId`. An unknown iroh remote fails closed (ADR-034 §3, Assumption 1 — no CA to fall back to). The iroh dial helper applies the same ADR-034 rule via iroh's own API; the consistency is in the rule, not in the type. ### `TlsError` The error type for `TlsServerConfig::new`, `TlsClientConfig::new`, and the `for_quinn()` accessors on both (`for_tcp_tls` is infallible — `TlsAcceptor::new` / `TlsConnector::new` cannot fail). A single `#[non_exhaustive]` enum with one variant per failure category, owned by `alknet-tls`. The shape, the rationale for single-enum-over-thin-wrapper, and the "what is NOT a variant" list are in [ADR-088](../../decisions/088-tlserror-shape.md); this section is the sketch. ```rust /// Errors produced by `TlsServerConfig::new`, `TlsClientConfig::new`, /// and the transport accessors (`for_quinn`; `for_tcp_tls` is /// infallible). One variant per failure category — match on the variant /// for the category, inspect the `#[source]` for the detail. #[derive(Debug, thiserror::Error)] #[non_exhaustive] pub enum TlsError { /// Cert or key file read / PEM parse. `io::Error` is the type the /// pemfile BufRead APIs return (pemfile funnels its own non-`Error` /// type into `io::Error` — see ADR-088 §"Gotchas" #2). #[error("loading cert/key material: {0}")] CertLoad(#[from] std::io::Error), /// Self-signed cert generation (rcgen). Server `SelfSigned` path. #[error("generating self-signed cert: {0}")] SelfSigned(#[from] rcgen::Error), /// rustls server or client config construction /// (`with_safe_default_protocol_versions`, `with_single_cert`, /// `CertifiedKey::from_der`, `RootCertStore::add`). Both paths. #[error("building rustls config: {0}")] Rustls(#[from] rustls::Error), /// `WebPkiServerVerifier::build()` — the unknown-X.509-remote /// client path (empty root store, invalid CRL). Distinct type and /// remediation from `Rustls`. #[error("building webpki verifier: {0}")] VerifierBuild(#[from] rustls::webpki::VerifierBuilderError), /// `QuicServerConfig::try_from(rustls::ServerConfig)` — the one /// path where `for_quinn()` fails. Distinct type /// (`NoInitialCipherSuite`, not `rustls::Error`); quinn-gated. #[cfg(feature = "quinn")] #[error("wrapping rustls config for quinn: {0}")] QuinnWrap(#[from] quinn::crypto::rustls::NoInitialCipherSuite), /// ACME config mismatch: "feature not enabled but `Acme` /// configured" (server), or "ACME identity is server-only; cannot /// be used for client auth" (client). A config error, not a /// wrapped third-party error. #[error("ACME configuration error: {0}")] AcmeConfig(String), } ``` **Scope boundary (ADR-088 §6).** `TlsError` is the **config-construction** error type — what `new` and `for_quinn` can fail on. Handshake-time errors (the unknown-raw-key fail-closed; a `rustls::Error::InvalidCertificate` from a rejected cert) are **handshake outcomes**, not config-construction errors — they flow through the transport's connector (`quinn::Endpoint::connect_with`, `TlsConnector::connect`), not through `TlsError`. ACME state-machine errors (`EventError`, `OrderError`, `CertParseError`) are stream events, logged in the spawned task, not `TlsError` variants — `new` spawns the state machine and returns immediately; the state machine's failures arrive asynchronously and are logged (ADR-082 §"Behavior-preservation invariants"). **Ownership.** `TlsError` lives in `alknet-tls`, owned by the crate that produces it. It is not re-exported from `alknet-core`; `EndpointError` is removed entirely after ADR-083 (both variants were vestigial), so core has no endpoint error type and does not need to know about `TlsError`. The assembly layer (hub/worker) depends on `alknet-tls` directly and gets `TlsError` from that dependency. ## Crate dependencies (in the dep graph) ``` alknet-tls ├── alknet-core (TlsIdentity, Ed25519SecretKey, fingerprint) alknet-core (loses TLS setup code + endpoint) ├── (rustls — only for fingerprint.rs types, if kept) alknet-call (pure protocol crate — no TLS/transport deps per ADR-089 §5) └── alknet-core (ProtocolHandler, Connection, types; ConnectionCredentials/ RemoteIdentity moved to core per ADR-091; CallCredentials stays in alknet-call) alknet-hub (multi-transport endpoint) ├── alknet-tls (TlsServerConfig — shared across quinn + TCP) ├── alknet-endpoint (AlknetEndpoint with quinn + iroh + tcp features, HandlerRegistry) ├── alknet-client (AlknetClient — outbound worker dials, ADR-089) ├── alknet-channels-call (ChannelClient) ├── alknet-call (CallAdapter, Dispatcher) ├── alknet-http (HttpAdapter) ├── alknet-core (Connection, ProtocolHandler, AuthContext, IdentityProvider) ``` `alknet-tls` depends on `alknet-core` only. No handler crate depends on `alknet-tls` — they depend on `alknet-core` for types and on `alknet-tls` only indirectly through the assembly layer. The assembly layer (the deployment binary) builds the `TlsServerConfig`(s), builds the transport endpoints (quinn/iroh/TCP+TLS), and hands them to `AlknetEndpoint` via `with_quinn` / `with_iroh` / `with_tcp_tls` (ADR-083). ## Design Decisions All design decisions are documented as ADRs in [decisions/](../../decisions/). | ADR | Decision | Summary | |-----|----------|---------| | [082](../../decisions/082-alknet-tls-extraction.md) | alknet-tls crate extraction | Extract TLS setup from alknet-core/endpoint.rs; `TlsServerConfig` shareable across quinn + TCP+TLS + iroh; one ACME state machine | | [083](../../decisions/083-endpoint-as-accept-loop-runner.md) | Endpoint as multi-transport accept-loop runner | `AlknetEndpoint` takes no TLS config; TCP+TLS is an owned transport (`with_tcp_tls`); `dispatch` public for SSH/WT; `acme-tls/1` guard moves to shared `dispatch` | | [084](../../decisions/084-aws-lc-rs-crypto-provider.md) | aws-lc-rs crypto provider | `rustls::crypto::aws_lc_rs::default_provider()` on all server + client config paths; matches iroh; FIPS-capable; do not switch to `ring` or process-default without a new ADR | | [086](../../decisions/086-endpoint-types-and-entry-points.md) | Endpoint types and entry points | Three endpoint types (web/native/iroh); split ALPN lists per endpoint type (resolves OQ-62); entry-point vs. endpoint ALPN distinction | | [087](../../decisions/087-tlsclientconfig-not-blocked-on-dial.md) | `TlsClientConfig` not blocked on dial seam | `alknet-tls` provides `TlsClientConfig` (client-side); not deferred behind OQ-55; breaks the circular hedge; hub-as-client is a first-class use case | | [088](../../decisions/088-tlserror-shape.md) | `TlsError` shape — single enum, owned by `alknet-tls` | Single `#[non_exhaustive]` enum, one variant per failure category (`CertLoad`, `SelfSigned`, `Rustls`, `VerifierBuild`, `QuinnWrap`, `AcmeConfig`); not a thin wrapper (the `for_quinn` failure is `NoInitialCipherSuite`, not `rustls::Error`); owned by `alknet-tls`, not re-exported from core | ## Open Questions See [open-questions.md](../../open-questions.md) for full details. - **OQ-59** (resolved): `fingerprint.rs` stays in `alknet-core`. The client-side `FingerprintPinVerifier` (now in `alknet-tls` per ADR-089 §5 — the original `alknet-call` → `alknet-tls` dep-edge concern that motivated keeping `fingerprint.rs` in core is dissolved). `fingerprint.rs` stays in core because `alknet-core`'s own `Identity`/fingerprint code uses it; `alknet-tls` re-exports. The `rustls` dep in core is narrow — production fingerprint code uses only `sha2` + manual DER; the `rustls::sign` usage is a test helper. - **OQ-60** (resolved): Where does transport construction live? The TCP+TLS accept loop lives in `alknet-endpoint` behind a `tcp` feature as an owned endpoint transport (`with_tcp_tls`). Builder functions are inlined by the assembly layer. See ADR-083. - **OQ-61** (dissolved): Multi-owner shutdown coordination. The problem does not arise — the endpoint owns all its accept loops (quinn, iroh, TCP+TLS); `shutdown()` stops them all. See ADR-083. - **OQ-62** (resolved): Does a hub pass the same ALPN list to both `TlsServerConfig`s? **Split list, by endpoint type** (ADR-086 §3). Each config advertises only the ALPNs its endpoint type's client class can negotiate — native ALPNs on the raw-key config, entry-point ALPNs + `alknet/channels` on the X.509/ACME config, native ALPNs on the iroh builder. The assembly layer filters `registry.alpn_strings()` per config. - **OQ-63** (resolved): `TlsError` shape — a single `#[non_exhaustive]` enum with one variant per failure category, owned by `alknet-tls` (not re-exported from core). Six variants: `CertLoad`, `SelfSigned`, `Rustls`, `VerifierBuild`, `QuinnWrap` (quinn-gated), `AcmeConfig`. The decision is grounded in the actual error-producing call sites and the dependency-crate sources; three findings drove the single-enum choice over a thin wrapper (the `for_quinn()` failure is `NoInitialCipherSuite`, not `rustls::Error`; `rustls_pemfile::Error` is not a `std::error::Error`; `WebPkiServerVerifier::build()` returns `VerifierBuilderError`). See [ADR-088](../../decisions/088-tlserror-shape.md) for the full rationale and the "what is NOT a variant" list. The `TlsError` sketch is in the [TlsError](#tlserror) section below. - **OQ-64** (resolved): `alknet-tls` provides `TlsClientConfig` (ADR-087). Not blocked on the dial-seam extraction — the TLS config is a prerequisite for the dial, not a consequence of it. Centralizes ADR-034 verifier selection + ADR-084 provider; the hub-as-client requirement makes it a prerequisite for the first hub deployment. The dial seam is now extracted as `alknet-client` (ADR-089, OQ-55 resolved); `TlsClientConfig` is consumed by `AlknetClient`'s QUIC and TCP+TLS dials. - **OQ-55** (resolved by ADR-089): `AlknetClient::dial()` — the transport-polymorphic dial seam. Extracted as a new crate `alknet-client` with three dial methods (`dial_quic` / `dial_tcp_tls` / `dial_iroh`). `TlsClientConfig` (OQ-64, resolved) is the prerequisite the dial consumes. See [`crates/client/README.md`](../client/README.md) and [ADR-089](../../decisions/089-alknetclient-native-dial-seam.md). ### Next session — client shape The client is now specced. [`crates/client/README.md`](../client/README.md) defines `AlknetClient` — the native client dial seam (ADR-089, resolves OQ-55). There are exactly two clients in the alknet client surface as far as `TlsClientConfig` and `AlknetClient` are concerned: **call** (`CallClient`) and **channels** (`ChannelClient`, a proxy over many ALPNs via channel 0). Both consume `TlsClientConfig` via the same three accessors (`for_quinn`, `for_tcp_tls`, `rustls_config`); iroh is the exception (shares the key, not the config). `AlknetClient` is the dial that feeds them — it produces a `Connection` and the protocol take-overs (`spawn_dispatch`, `from_connection`) consume it. The per-protocol QUIC convenience constructors (`CallClient::connect` / `ChannelClient::connect_quic`) are **removed** per ADR-089 §5 — the dial is centralized in `AlknetClient`, and the protocol crates shed their TLS/transport deps. The `alknet/register` ALPN (native registration entry point, parallel to HTTP registration in OQ-58) is named by ADR-089; its wire protocol is deferred (OQ-66). ## References - `docs/architecture/decisions/027-tls-identity-redesign-acme-rawkey-decoupling.md` — `TlsIdentity` (RawKey / X509 / Acme), RFC 7250, browser limitation - `docs/architecture/decisions/030-peerentry-and-identity-id-decoupling.md` §6 — fingerprint normalization (`ed25519:` across quinn/iroh) - `docs/architecture/decisions/034-outgoing-only-x509-and-three-peer-roles.md` — client-side verifier selection (CA vs fingerprint pin) - `docs/architecture/decisions/065-connection-from-stream-generic-single-stream.md` — `Connection::from_stream`/`from_bidi` (TCP+TLS path) - `docs/architecture/decisions/010-alpn-router-and-endpoint.md` Amendment 2 — TCP+TLS is a first-class owned transport (`with_tcp_tls`); supersedes Amendment 1's sibling-loop framing - `docs/architecture/decisions/086-endpoint-types-and-entry-points.md` — three endpoint types (web/native/iroh); split ALPN lists per endpoint type (resolves OQ-62); entry-point vs. endpoint distinction - `docs/architecture/decisions/087-tlsclientconfig-not-blocked-on-dial.md` — `TlsClientConfig` (client-side); not blocked on the dial seam; breaks the circular hedge; hub-as-client requirement - `docs/architecture/crates/endpoint/README.md` — `AlknetEndpoint` (the endpoint spec; TLS config is built by `alknet-tls`, not the endpoint — per ADR-083) - `docs/architecture/crates/core/config.md` — `TlsIdentity`, `StaticConfig` - `crates/alknet-core/src/endpoint.rs` — the server-side code being extracted (`build_rustls_server_config`, `TlsSetup`, `RawKeyCertResolver`, `Ed25519SigningKey`, `AcceptAnyCertVerifier`, `generate_self_signed_cert`, `load_cert_chain`, `load_private_key`) - `crates/alknet-call/src/client/call_client.rs` — the client-side code being extracted (`build_quinn_client_config`, `build_client_auth`, `select_server_verifier`, `load_platform_root_cert_store`, `FingerprintPinVerifier`, `RawKeyClientCertResolver`, `NoClientCertResolver`, `Ed25519SigningKey` (duplicate), `load_cert_chain`/`load_private_key` (duplicates)) - `crates/alknet-core/src/fingerprint.rs` — fingerprint extraction (shared by server endpoint and client verifier)