dBFT and its way through science
After many months on hard work and heat discussions, a history of dBFT has been published online at journal Future Internet, hosted at the Multidisciplinary Digital Publishing Institute (MDPI). The MDPI focuses on Open-Access research, with much faster peer-review processes than traditional academic journals. The paper had participation from researchers and active voices from distinct Neo community members, including NeoResearch and NNT, from academia (Brazilian universities UFF and UERJ) and from NGD research team.
The paper entitled “Challenges of PBFT-Inspired Consensus for Blockchain and Enhancements over Neo dBFT” presents theoretical assumptions on peer-to-peer for efficient blockchain consensus, and for decentralized ledger technologies on general. The article describes how dBFT tackles challenging Byzantine Fault Tolerance problems, inspired by PBFT (Castro&Liskov, 1999) and other groundbreaking works from literature. However, porting these ideas from single-transaction systems to blockchains is not straightforward. Since its inception, dBFT 1.0 has efficiently solved hard block proposal issues over an asynchronous and decentralized P2P network, and over a non-deterministic memory pool. Some of these challenges required adjustments on dBFT 1.0, giving birth to dBFT 2.0, seeking to remove multiple block proposals over a One Block Finality network (known as sporks). The paper describes how dBFT 2.0 differentiates from dBFT 1.0, and suggests further improvements to be made in the future (proposals named dBFT 2.0+).
One of the key points in the article is related to mechanisms that provide multiple simultaneous block proposals, which authors argue could improve performance and robustness of the consensus (specially when dealing with systematically crashes on primary dBFT nodes). This mechanism was named dBFT 3.0, as a proposal to be discussed by Neo community members, towards some future adoption on practice. Finally, some extra discussions indicate other possible enhancements, named dBFT 3.0+, that could bring some extra guarantees to the decision-making by consensus nodes, but also bring some more burden in peer-to-peer network.
The publication of this paper in a academic journal was a victory to both Neo community and the researchers involved, specially due to the peer-review system adopted by scientific journals. This means that the publication is only accepted after anonymous referees criticize the work, require revisions, and only if experts are satisfied with responses and changes, publication is recommended. This work was built over a public report by NeoResearch community, including extra changes during publication process at MDPI. The article is co-authored by Igor M. Coelho, Vitor N. Coelho, Rodolfo P. Araujo, Wang Yong Qiang and Brett D. Rhodes.