Otter

worker threads — one system thread can manage thousands of engines. Each

Process = one Runtime + Context + mailbox. Never shares JS state.

Scheduler = M worker threads pulling processes off a shared run queue.

Every file on the command line starts as its own process (pids 0..n ); the

runtime exits when all processes (including spawned ones) have finished.

otter can expose a JSON-RPC 2.0 control socket over TCP, letting an external

kill , send , rename , count , and shutdown . Enable it with

Serialize value to JSON and deliver it to pid 's mailbox. Messages to dead pids are dropped silently.

Suspend until a message arrives; resolves with the parsed value. With a timeoutMs it rejects with a TimeoutError if no message arrives in time (a message that races the deadline stays in the mailbox).

Suspend the process for at least ms milliseconds; does not touch the mailbox.

Voluntarily give up the current slice and rejoin the back of the run queue. (Not named yield because that word is reserved inside async-function bodies, which is how top-level-await scripts are parsed.)

Request termination of a live process; returns true if pid is live. Best-effort: the process is reaped at its next scheduling boundary (parked/sleeping processes immediately). Killing an unknown pid returns false .

Rename the current process; visible in listProcesses() / processInfo() .