Different Perspectives on FLP Impossibility

We demonstrate possibility for vector consensus under the model used by Fischer, Lynch, and Patterson (FLP) to prove impossibility of binary consensus - in full asynchrony and up to one unannounced faulty process. We also show that: i) figuring out a vector dataset is an inevitable phase of binary agreement; and ii) agreeing on a vector dataset is sufficient for binary agreement. Key finding: the FLP result proves impossibility the processes to always produce binary value from a figured-out dataset that contains less than the entire initial state, which causes the impossibility to terminate.