Fairness takes three forms. The definition for fairness for our consensus engine means:

  • fair access: meaning all peers are equally able to submit messages into the network without the risk of any one peer preventing the propagation of the submitted message. Even if malicious peers attempt to hinder a specific message, the gossip protocol is configured in a way to circumvent the blockage.

  • fair order: meaning that the order of a message in the final sequence is mathematically determined by an unbiased distribution of messages into the network. Message ordering is not controlled by, nor can it be significantly affected by, any one peer’s effort. If all messages are ordered as a result of being witnessed by 2/3 of the network peers, then any subset of peers less than 1/3 have no effect on the outcome of ordering.

  • fair timestamp: meaning that in the ordering process, the implementation may calculate a median timestamp for each message submitted into the network. This timestamp may be the median timestamp of when 2/3 peers in the network become aware of the message in question, or it may be a deterministically derived timestamp applied to the order of messages. Either way, any timestamp that is generated by consensus will be consistent between peers and be equally fair as with ordering.

Last updated