Peers should be banned when a request fails or when they misbehave

[tomaka]

When a networking request fails, or when it succeeds but the result is invalid, the peer needs to be disconnected from and banned.

I thought I had an issue open about this, but I can't find it.

For some aspects this is easy, but for some, such as an invalid justification, this is harder due to the asynchronousness of communications between services.

[tomaka]

I don't think that the network services should automatically ban peers due to failed requests. The network services always send requests only because the sync/consensus service called a function specifically to send a request. Banning a peer as a result feels like a hidden side effect.

Instead, the networking service should expose the concept of banning on its public API, letting the sync/consensus service ban peers.

When the sync/consensus service detect that a peer has misbehaved, it should be immediately removed from the list of sources, and the networking service then bans it. In the meanwhile, any message concerning this peer is ignored.

[tomaka]

When the sync/consensus service detect that a peer has misbehaved, it should be immediately removed from the list of sources, and the networking service then bans it. In the meanwhile, any message concerning this peer is ignored.

I don't think we actually need to do that. It would add a lot of complexity, and isn't strictly necessary.