koi-demo

[sayertindall]

No description provided.

[lukvmil]

I think we'll need to discuss some of this over a call, I'll make a few notes here first:

• I appreciate you adding sanity checks for config variables and connection failures, there's definitely a lot of the core code that needs to be made more robust in koi-net/rid-lib.
• Not going to look into the implementation details for the upgraded backfill, but it looks like it'll be much more robust than the quick solution I had prev :)
• The HackMD state JSON can be wrapped up into the config.yaml, see the implementation in the slack-sensor-node for your reference. It has a config section which includes a latest timestamp field, which can be regularly updating by calling node.config.save_to_yaml().
• The coordinator contact handler probably shouldn't be rewritten, we want standard behavior across all nodes for this action. I left this comment in a previous review, nodes should not propose that other nodes subscribe to them. Only edge targets should be proposing edges. The coordinator will decide by itself to subscribe to KOI-net node/edge events from the hackmd-sensor when they do a handshake. Finally, not sure why the coordinator is setup to receive HackMDNote RIDs, it should only be concerned with nodes and edges.
• The extended behavior in the lifespan function seems redundant, if there's issues with the default behavior lets discuss changes to the koi-net library. For example, the first contact will be automatically contacted after running node.start() when the node doesn't know anyone else. If this behavior is failing lets try to get a general fix.