In this guide
Anticipation and publication are different statesA held media request differs from a held playlist requestPreserve the actual published resultAn HLS preload hint identifies a resource or byte range expected to be needed soon. In Low-Latency HLS, a client can request hinted media before it is fully available and the server can respond when the media is ready.
The hint reduces an avoidable discovery/request round trip. It does not mean the resource is already a finalized playable timeline entry or that every anticipated segmentation decision is permanent.
Anticipation and publication are different states
Apple documents that a hinted partial segment may later appear with changed surrounding context and that planned segmentation can change. The client must use the actual playlist/media rules rather than treating a hint as an unconditional promise about future content.
An original hypothetical hint anticipates a part after the current boundary. The final published timeline introduces new initialization context before that part. A client that ignores the updated context could interpret the media incorrectly even though it requested the hinted URL early.
A held media request differs from a held playlist request
Blocking reload waits for a newer playlist state. Preload retrieval waits for an anticipated media or initialization resource. The two can be pending together and become useful around the same production event. Their resource roles should remain separate in diagnostics.
| Request | Waits for |
|---|---|
| Blocking playlist reload | A requested newer playlist state |
| Hinted media request | The anticipated resource becoming available |
| Ordinary listed-part request | An already described timeline resource |
| Initialization request | Required media interpretation context |
Preserve the actual published result
For a failure, record the hint, its request outcome and the later playlist entry/context where available. A changed plan is different from a currently listed resource returning an unexpected failure. Do not infer permanent absence from one anticipated request alone.
Viewer-side URL editing cannot make the publisher produce a hinted part sooner. Use supported clients and a minimal redacted trace. The useful conclusion concerns actual availability and context, not a fixed latency guarantee.