wondertan / go-ipfs-recovery Goto Github PK

Background

Currently, recovery goes after IPFS says that it can't find any more blocks on the network, but this is not convenient from UX standpoint. That pushes a user to wait minutes even if there are enough nodes for recovery.

Goal

We need to define the actual point where IPFS will fallback on to recovery without degrading UX.

Solutions

• Recover every time we see there are enough nodes.
• Recover at the end of a timeout started when a node found min amount of blocks.

Questions

Recover often abusing CPU but serve content to a user faster, or recover less but keep the user waiting for content on the network more?

When restoring a node, we need to iterate recursively to leaf nodes, create redundancies for them, then update the link in the parent.