Reply timeout or oversized response: transport limits do not prove no write
Byte/time limits protect the consumer; they are not business results or response SLAs
Native reply bodies are capped at 20000 bytes. Non-history responses are capped at 40000 bytes; history at 128 KiB, including streaming enforcement. Request/resource timeouts vary by operation. Hitting limits does not establish that nothing was recorded server-side and cannot automatically authorize resending. No request, new API or tuning feature is supplied.
Different from URL, identity and retrieval review
URL preflight checks shape, request mapping checks identity/receipts, and retrieval checks coverage. Here why client transport settings may terminate reading and what that proves matters. Original models, identity and target checks remain.
| Check | Verified scope | Manual review | Not established |
|---|---|---|---|
| Request body | Up to 20000 bytes, including JSON metadata | Encoded total, not text length | 20000 characters always deliverable |
| Non-history response | Up to 40000 bytes | Actual response/decoding | HTTP success means business success |
| History response | Up to 128 KiB | Stream count/model validation | All history in one response |
| Content-Length | Oversize rejected early; stream still bounded | Unknown/untrusted length versus actual bytes | No header means unlimited reading |
| Request timeout | History 45 seconds, others 25 | Client configuration/actual error | Service completion deadline promised |
| Resource timeout | History 60 seconds, others 45 | Separate resource budget | End-to-end handling SLA |
| Session/redirect | Native cookie/credential stores separate; no automatic redirect | Original source/actual authority | Browser login becomes native write access |
| Uncertain submit | Original pending/check path remains | Request identity/real outcome | Failed response read means no effects |
Synthetic transport note
Known: response reading timed out after a submit. Unknown: whether the reply was recorded. Preserve pending and use original check; do not invent absent or blindly change URLs/request IDs.
Bytes/times are source settings, not load, uptime, network-latency or cost evidence. Page login/native authority cannot be inferred from each other. Source/epoch/version/confirmation guards remain.
FAQ
Does 20k body mean 20k CJK characters?
No. It counts JSON-encoded bytes including metadata, not a direct character allowance.
Does an oversized response prove no write?
No. Reading failure differs from server outcome; check the original request.
Is reading unlimited without Content-Length?
No. Streaming enforces byte limits too.
Do 45/60 seconds guarantee completion?
No. They are separate client configurations, not business SLAs.
Can redirects automatically continue on another site?
The native path refuses automatic redirects; compatibility does not relax source/access boundaries.
Was an endpoint/load test run?
No. Fixed source only, without actual requests, performance or availability acceptance.
Reading and source
Feed URL · Request/receipt mapping · History coverage · Start · Privacy
