Server lifecycle

Create and release a server

Server:new(port) binds all IPv4 interfaces. Use new_on(port, address) for an explicit bind address. new_on_with_limits also accepts an owned limits record. A null server means initialization failed; no server handle may be used afterward.

import http
let serve_one_request = fn () -> i64 {
    let limits = Http:Limits:defaults()
    if !limits { return 1 }
    limits.body = 65536
    let server = Http:Server:new_on_with_limits(8080, "127.0.0.1", limits)
    if !server { return 2 }
    defer server.destroy()
    if !server.route("GET", "/health", fn (request:Http:Request*, response:Http:Response*) -> void {
        response.text(200, "ok")
    }) { return 3 }
    return !server.serve_once()
}
// Calling this waits for one client connection.

Choose synchronous or worker execution

serve_once is synchronous and returns 1 for a served connection, 0 on failure. run starts the bounded worker loop and blocks. Do not call destroy concurrently with active workers: destruction requires that serving work has already stopped. The current public API does not provide a graceful stop token or promise cancellation through destruction.

Choose a transport boundary

The library implements HTTP transport, not TLS termination. Select a deployment arrangement appropriate to your application. Protocol parsing handles framing; route handlers still validate application data and authorization.