Network protocol keeps requesting data the peer does not have

[gerardsn]

Situation:

• Alice tries to handleTransactionList received from Bob
• While adding one of the transactions to the DAG, it fails with dag.ErrPreviousTransactionMissing
• Alice immediately sends a State message to Bob to restart the set reconciliation
• Bob does not have the missing transaction so ends up sending the same transactionList as before (DB is corrupt? attacker?)
• go back to step 1

It's unlikely that Bob is going to send a different result than last time, so it's likely to fail again.

This generates a lot of traffic since there is no delay between the two attempts. A solution is not obvious to me since adding a delay is just reducing the amount of spam a bit, but permanently blocking is not an option either.

[woutslakhorst]

It might be best to stop processing when a transaction is missing? Most likely you'll just get it from a different node.

[gerardsn]

I agree. This is a very specific example, but requesting the same data again is unlikely to result in a different outcome. The only plausible situation I can think of is when a DID needed to validate a transaction has been added to the DAG, but not yet processed by the VDR.

Stopping the conversation/reconciliation adds a retry interval equal to the length of the gossip interval.

[woutslakhorst]

We'll need to change RFC017 first...
nuts-foundation/nuts-specification#262