docs: recommend bounded retry/backoff for token exchange failures
Update integration guidance to avoid tight retry loops during transient token endpoint failures by using bounded backoff strategies.
This commit is contained in:
@@ -140,6 +140,7 @@ try {
|
|||||||
- Avoid printing tokens in production pages.
|
- Avoid printing tokens in production pages.
|
||||||
- Expect `429 Too Many Requests` from provider control endpoints under abuse protection; implement backoff/retry instead of tight loops.
|
- Expect `429 Too Many Requests` from provider control endpoints under abuse protection; implement backoff/retry instead of tight loops.
|
||||||
- Apply the same backoff strategy for provider `userinfo` calls that return `429 Too Many Requests`.
|
- Apply the same backoff strategy for provider `userinfo` calls that return `429 Too Many Requests`.
|
||||||
|
- For token exchange and refresh, apply bounded retry/backoff on transient provider failures instead of immediate repeated retries.
|
||||||
- If you call provider WebAuthn login challenge/verify endpoints directly, use `POST` only.
|
- If you call provider WebAuthn login challenge/verify endpoints directly, use `POST` only.
|
||||||
- Configure OAuth client redirect URIs as HTTPS in production; reserve HTTP for localhost loopback testing only.
|
- Configure OAuth client redirect URIs as HTTPS in production; reserve HTTP for localhost loopback testing only.
|
||||||
- Rotate client secrets for confidential clients.
|
- Rotate client secrets for confidential clients.
|
||||||
|
|||||||
Reference in New Issue
Block a user