Skip to main content

process_request

Function process_request 

Source
pub async fn process_request<S>(
    raw_bytes: &[u8],
    service: &S,
    verification: EnvelopeVerification<'_>,
    signing_key: &SigningKey,
    nonce_cache: &InMemoryNonceCache,
    carrier: CarrierContext,
) -> Result<Vec<u8>>
where S: RequestService,
Expand description

Process a request through the full envelope verification pipeline.

Unified handler for all transport front-ends. The only difference between them is envelope signer verification, controlled by verification.

§Pipeline

  1. Unwrap SignedEnvelope and verify the signature (mode-dependent), under the process-global verify policy + kid-anchored PQ trust store.
  2. Verify JWT claims (sub, exp, aud, scope, downgrade protection).
  3. Dispatch to service.handle_request() with a verified EnvelopeContext.
  4. Sign the response with the server’s signing_key.

§Streaming

A streaming handler returns a Continuation (the server-side streaming response that runs after the reply). As of #186 that task is spawned here, via crate::streaming::spawn_streaming_response, rather than handed back to the transport front-end — so every front-end (ZMQ RequestLoop, WebTransport server, generic-plane LocalServiceBridge) is uniform “bytes in → bytes out” and the generic plane no longer has to spawn (or, worse, reject) continuations itself. Invariant: process_request must therefore run on a tokio::task::LocalSet (it already did — RequestService is ?Send); the spawned task is ?Send.

§Returns

  • Ok(response_bytes) - Signed response. Any streaming pump has already been spawned onto the current LocalSet.
  • Err(e) - Processing error (already logged)