- file domains are created, queried, sized and deleted by the file service
(POST/GET/DELETE /file/v1/domains), with per-domain quota enforced on write
- /file/v1/options publishes the quota choices; nothing about files lives on
the host any more
- tickets carry the issuing space and a unique id: a domain only accepts tickets
from its own space, and read/delete tickets work once (no replaying a counted
download)
- uploads stay pending until the client confirms them; abandoned ones are swept
after 24h, and an upload ticket can cancel its own bytes
- backup target is S3 (SigV4 client, no SDK) instead of WebDAV
- CORS allows the methods the browser really uses (POST, DELETE) and exposes Retry-After
- new projects WebDrop.FileApi (file service), WebDrop.FileProtocol (shared tickets and receipts) and WebDrop.Sync (WebDAV)
- file content of any size goes to the file service; chat, avatars and space structure stay on the main backend
- tickets issued by the main backend, receipts by the file service, deletion queue and daily reconciliation
- daily database snapshot and file sync to external storage (OpenList/WebDAV) with a 30 day retention
- reorganize Core, Storage and Api into feature folders (namespaces unchanged)
- web client uploads file frames to the file service and reads them with tickets
justme