Dispute distribution guide

[eskimor]

Note: Guide PR needs update. The following things have changed:

There is no separate message for reporting unavailable candidates anymore, instead ImportStatements got a more informative back channel:

		ImportStatements {
		/// The hash of the candidate.
		candidate_hash: CandidateHash,
		/// The candidate receipt itself.
		candidate_receipt: CandidateReceipt,
		/// The session the candidate appears in.
		session: SessionIndex,
		/// Statements, with signatures checked, by validators participating in disputes.
		///
		/// The validator index passed alongside each statement should correspond to the index
		/// of the validator in the set.
		statements: Vec<(SignedDisputeStatement, ValidatorIndex)>,
		/// Inform the requester once we finished importing.
		///
		/// This is, we either discarded the votes, just record them because we
		/// casted our vote already or recovered availability for the candidate
		/// successfully.
		pending_confirmation: oneshot::Sender<ImportStatementsResult>
	},

with ImportStatementsResult looking like this:

pub enum ImportStatementsResult {
	/// Import was invalid (candidate was not available)  and the sending peer should get banned.
	InvalidImport,
	/// Import was valid and can be confirmed to peer.
	ValidImport
}

Details in implementation PR.

[rphmeier]

What are validator operators meant to do in this situation?

[eskimor]

On a live network - check their internet connection, on a test network: find the bug. Like we had the case already a couple of times that we were not connected properly and we did only realize because of other issues that have been caused by this. Dispute distribution already tries its best - it will try to send requests, even if the receiver is not connected to us and it will keep trying.

I was just thinking about disputes and because they are so critical, I wanted to do whatever I can to ensure our messages gets out, but general connection warnings should go to gossip support already, I guess. And warnings when a dispute is already happening are a bit late - still useful though as a additional safety guard though. If validators become aware of a dispute that did not work out for some weird reasons/bugs/whatever we still have governance - better than nobody noticing.

[rphmeier]

Looks good modulo nits. I think the spam-resilience strategy makes sense and should protect against any practical attacks.

[rphmeier]

Follow-up changes should go in another PR. Thanks!