rate_limit.nx

rate_limit.nx -- token-bucket rate limiter. Standard algorithm (RFC 2698 / IEEE; also what HAProxy, nginx, AWS API Gateway, and Stripe use): - Bucket has a max capacity (burst size) of B tokens - Tokens refill at rate R per second up to B - Each request consumes 1 token (or more for weighted ops) - Request allowed iff current tokens >= needed; otherwise reject (429 Too Many Requests) or wait Time is passed in as unix_microseconds -- caller reads a monotonic clock and hands it to us. This keeps the module pure + testable and decoupled from the OS time source. For real use pair with ntp.nx + sys_clock_monotonic. Use cases: HTTP API throttling (per IP / per token), login attempt rate limiting (anti-brute-force), outbound API call pacing, mail-send pacing, git clone pacing. Invariants: RL1 tokens clamped to capacity on every refill (bucket never overflows). RL2 Time going backwards does nothing (malicious / NTP adjustment) -- we record max(now, last_refill). RL3 Rate is expressed in tokens per second; sub-second granularity handled via microsecond timestamps.