The HoneyBadger tests currently contain an oddly specific "faulty shares" test: Should that be moved to ThresholdDecrypt tests instead? (See #400.)
Apart from that, we need to ensure that the RandomAdversary doesn't almost exclusively send completely nonsensical messages (e.g. with huge epoch numbers), but uses random distributions that are more likely to uncover bugs.
Let's also evaluate whether additional message ordering strategies would make sense (FIFO, LIFO, different "node speed" distributions)—always with the restriction that all messages must eventually be delivered, of course.
The
HoneyBadgertests currently contain an oddly specific "faulty shares" test: Should that be moved toThresholdDecrypttests instead? (See #400.)Apart from that, we need to ensure that the
RandomAdversarydoesn't almost exclusively send completely nonsensical messages (e.g. with huge epoch numbers), but uses random distributions that are more likely to uncover bugs.Let's also evaluate whether additional message ordering strategies would make sense (FIFO, LIFO, different "node speed" distributions)—always with the restriction that all messages must eventually be delivered, of course.