Tsampi BFT: Leaderless One-Round Voting with Parameterized Finality

Tsampi BFT introduces a leaderless approach to Byzantine fault-tolerant state replication, eliminating the need for a designated leader in each round. By consolidating voting into a single round per proposed block, the protocol reduces the number of message exchanges that typically accompany multi-round consensus. The dual-linked vote chain is a novel data structure that simultaneously tracks the lineage of blocks and the order in which validators endorse them. This change shifts the design focus from leader election mechanics to the maintenance of linked vote metadata.

Adopting Tsampi BFT imposes additional storage and processing requirements on validators. Each validator must maintain the dual-linked chain, which stores two pointers per vote to capture both block causality and endorsement order. Processing a single vote round still requires validators to broadcast and collect votes from a quorum, but the reduced round count lowers latency at the cost of more complex bookkeeping. Engineers must ensure that the implementation correctly updates both links to preserve the exact-lineage property.

The protocol’s safety guarantees depend on the underlying fault-tolerance threshold and network timing. If message delays exceed the window needed to establish exact-lineage finality in two blocks, finality may be delayed, requiring more blocks to achieve certainty. Should the number of faulty validators exceed the protocol’s resilience bound, safety can be compromised regardless of the vote-round optimization. The parameterized finality feature allows tuning, but setting it too low under adverse conditions can lead to unsafe outcomes.

For engineers building or operating replicated services, Tsampi BFT offers a clear trade-off: lower latency and simpler leader handling versus increased complexity in vote-chain maintenance. The protocol is most beneficial in environments with reliable, low-latency networks where the two-block finality target is consistently attainable. In highly volatile or partitioned settings, the added overhead of managing dual links may outweigh the latency gains, and fallback to more traditional multi-round BFT designs might be preferable.