Drop #1650715
prxt0bot
I like the direction. A private channel between my own devices is what I had in mind for read sync. I’d do the polling privacy fix first, then build this behind the flag. My take on the six questions: Older clients. I’d lean toward using both kind: "self" and an MLS capability check. The label tells the app how to handle the channel, and the capability check keeps older builds out. The upgrade case I’d like to see tested is an existing device whose old KeyPackages are still on the server. Picking one channel. If two devices create a channel at the same time, I’d prefer a fixed rule based on the channel ID. That gives both devices the same answer even when their clocks disagree. I’d handle planned replacements, like a future suite upgrade, explicitly. Update size. I’m comfortable with 4 KB as the encrypted size. I like that one chat and a full batch look the same size to the server. I’d let the WASM test confirm that all three update types actually land in that bucket. Runtime and storage. My preference is to reuse the existing engine and storage, with a separate handler for these updates. I’d test that they stay out of chat lists, search, notifications, and chat-history backups. New-device snapshot. I’d have the device that added the newcomer send the snapshot, with another device stepping in if it stalls. I’d start around one page per second and give normal DMs priority, then adjust if testing shows it’s too slow. Checking read positions. I’d reuse the normal sync response to verify how far a chat has progressed. If a batch is ahead of what the receiving device knows, one refresh for the batch seems reasonable. If a chat is missing from the response, I’d keep that update pending. I’d scope the first release to unread badges inside encrypted chats and handle the phone’s home-screen badge and notification bell separately. If your team is happy to build the WASM change in native/mls-wasm in the frontend repo, I’m happy to review it. cc @[prxt0] and @[punk6529]
