HTTP 429 Retry-After: Delay Seconds vs HTTP Dates

Interpret two hypothetical timing values from a stated receipt time, then separate the waiting.

Unbranded physical desk timer and closed midnight-blue laptop on a pale blush desk
AI-generated illustration · Conceptual timing scene; not an API capture, tested retry, specific waiting duration or successful request.
On this page
  1. Two time references
  2. Timing and retry behavior
  3. Identify the form before you calculate
  4. One value does not prove a quota
  5. The next response to inspect

Two time references

Illustrative Retry-After flow: delay-seconds means waiting a duration; HTTP-date names a time; retry does not guarantee success
AI-assisted image · Simplified timing explanation based on MDN and RFC 6585; not an actual API log, service policy or automatic-retry instruction.

Both examples below are hypothetical. Assume receipt at 12:00:00 GMT on 26 October 2026. No response was captured and no retry was performed.

Delay-seconds example: Retry-After: 120 gives an interval ending at 12:02:00 GMT. The calculation is receipt time plus 120 seconds, not request-start time plus 120 seconds. This form uses a whole-number value of zero or greater.

HTTP-date example: Retry-After: Mon, 26 Oct 2026 12:03:00 GMT names 12:03:00 GMT. With the assumed receipt time, that is 180 seconds later. Unlike the first form, the value names a time rather than a duration. The two forms follow MDN syntax and directives; the arithmetic is illustrative.

Timing and retry behavior

RFC 6585 does not prescribe one universal request-counting scope or user-identification method. Do not turn a waiting value into an assumed quota. See RFC 6585, section 4.

For practical review, keep three questions separate: what time the header indicates, whether the client will retry, and what happened after an actual attempt. Confirm client behavior and service limits in their applicable documentation rather than treating the diagram as an implementation specification.

Identify the form before you calculate

The two examples in this article share the same receipt time — 12:00:00 GMT on 26 October 2026 — and still point to different moments, because the form decides what the value means. Delay-seconds of 120 ends the wait at 12:02:00 GMT: receipt time plus 120 seconds, not request-start time plus 120 seconds. An HTTP date of Mon, 26 Oct 2026 12:03:00 GMT names 12:03:00 GMT, which is 180 seconds after the same receipt time. So the first step is never the arithmetic. Record the header value and the receipt time, identify which form you are looking at, and only then calculate. A value read as the wrong form gives a confident, wrong answer.

One value does not prove a quota

RFC 6585 does not prescribe one universal request-counting scope or user-identification method, and this article is explicit about what follows: do not turn a waiting value into an assumed quota. A Retry-After of 120 seconds tells you how long to wait before the next attempt; it says nothing about how many attempts you are allowed, how the service counts them, or how it identifies you. Keep the three questions in this article separate — what time the header indicates, whether the client will retry, and what happened after an actual attempt — and confirm client behavior and service limits in their applicable documentation. The header is a waiting instruction, not a rate-limit policy.

The next response to inspect

Use one response you control. Record its header value and receipt time, then identify the form before calculating. If the value is missing or unclear, record that uncertainty instead of inventing a delay. Leave the outcome as not observed until an actual attempt supplies evidence.

Source basis: supplied documentation captures dated 25 October 2026; no new source check or live availability check was performed. Refresh this explanation if the documented semantics or applicable client/service instructions change.

Part of the API Security Review Hub: eight practical guides for reviewing AI automations that call APIs.

Comments

No comments yet. Why don’t you start the discussion?

    Leave a Reply

    Your email address will not be published. Required fields are marked *