Composable DHT Open Questions

[dennis-tra]

During our Paris meeting, we went through the composable DHT design page and had a couple of questions.

• Do we really need to learn the feature preference list of the remote peer? We don't think so
• Couldn't the identify protocol serve as a feature discovery? -> it would be great if we could avoid inventing a new protocol for feature discovery
• Generally, the Feature/Protocol search must be clarified
  • Currently, FIND_NODE requests will return any nodes in the routing table of the remote peer and not necessarily the ones that support the feature that I'm searching for
  • sparsly populated features will be hard to find
• This could be solved with two not-really satisfying ways. Add a new FIND_NODE_WITH_FEATURE protocol. Connect to all peers that get returned from a FIND_NODE RPC.