refactor Addressbook proposal

The main process is horribly complex.

Addressbook is a class in the main process that discovers and keeps a stateful collection of node addresses. A node is chosen randomly from the pool when starting Gaia Lite.

The usage of the address book is essentially as follows:

1. Load seed nodes from the network’s config.toml file.
2. Load more nodes that were saved to disk during the previous session.
3. Filter out nodes that don’t respond to a simple query.
4. Discover new nodes by asking a node for its peers.
5. Save nodes to disk for the next session.

I propose that we replace this complex, stateful class with a single stateless function. The function will:

• take a set of seed nodes, discover new ones, and return a combined set of nodes known to respond
• combine the aliveness query with peer discovery to reduce the amount of network traffic
• make the network queries in parallel for increased performance
• not talk to the disk to eliminate the need to mock file system calls in tests

[NodeGuy]

I agree we’ll probably want to have it again and I don’t see any harm in leaving it as it is since it can handle the case of a single node already.

[faboweb]

Right. So should we archive the addressbook code then? In the future we probably want to have this again.