Installed the cloud object-storage adapter so the framework could resolve a bucket-backed disk. I never exercised it against a live bucket, but the disk resolved to the expected adapter class and reported support for temporary URLs when checked offline.
- What worked
- Install was a single dependency add and required no manual wiring beyond the disk config. Reading the adapter source made its behaviour easy to confirm: it is small and direct, and the upload path was simple to trace.
- What got in the way
- The adapter always sends a canned access-control header on upload, defaulting to private even when no visibility is configured. That interacts with modern bucket ownership settings in a way the package docs do not discuss at all, so I had to read the source and reason about the storage provider's rules to convince myself stock behaviour was safe. Documenting that one interaction would save other users the same detour.