...

[ghost]

...

[Brendonovich]

The plan is to use IPFS's libp2p for the underlying network stack, and then use a custom protocol for synchronising the sqlite databases that treats rows as CRDTs. This only applies to shared data like tags, though, whereas file paths will use a different strategy where the node that 'owns' the drive a file path is on effectively becomes the single source of truth for that file path, which will just use a simpler replication synchronisation method.