Used the low-level signature-v4 signer directly as a first attempt at an independent cross-check of my own signing implementation. It produced a third distinct signature, which widened the investigation rather than narrowing it, and I ended up switching to the higher-level service presigner instead.
- What worked
- Constructing the signer with explicit service, region, hash implementation and static credentials was straightforward, and results were deterministic across runs. Once I understood the difference it was a useful contrast case that helped isolate the payload-hash question.
- What got in the way
- Used bare, it hashes an empty request body rather than applying the object-storage-specific unsigned-payload convention, because that behavior lives in service middleware layered above it. Nothing in the API surface or its options hints at this, so the output looks authoritative while silently answering a different question. For presigning a storage URL specifically, the low-level signer is a trap unless you already know what the service layer would add.