The resurgence of distributed systems computing surely owes a thing or two to the pseudonymous Satoshi Nakamoto. Whoever Satoshi actually is has become, perhaps by Nakamoto's own design, less important than what Satoshi left us – the consensus methodology known as proof-of-work (PoW). Now, some new mysterious entities have provided us with what Emin Gün Sirer – Ethereum whiz, Cornell University associate professor, and co-director of the IC3 Initiative for Cryptocurrencies and Smart Contracts in New York City – is crediting as a groundbreaking new set of consensus protocols.
New Alternatives to Building a Blockchain
Although PoW is far younger than classical distributed consensus, it seems to be the most famous of distributed systems consensus algorithms. PoW is what powers and secures bitcoin. It is one of the elemental sparks that helped usher blockchain technology into existence.
But there are alternatives to PoW. One of the more well-known is termed proof-of-stake (PoS), which dispenses with the computational costs of PoW in favor of requiring a node to "stake" its money in order to participate in consensus. Staking more or less value allows participants to indicate their faith in a given block of a developing blockchain; it also dispenses with the huge energy costs of mining and the associated environmental impact, as well as the budgetary costs of advanced mining rigs.
Then, last week, an unknown team of researchers sent an email to the InterPlanetary File System outlining what appeared to be a new consensus family. The anonymous team called its family of protocols Snowflake, Snowball, and Avalanche.
What Makes This Snowfall Different?
While the initial shock of the anonymous whitepaper, written under the authorship of the pseudonymous "Team Rocket," caused some prominent distributed systems researchers to question the security of this new family – ETHNews had the opportunity to discuss Team Rocket's whitepaper with Sirer, one of its stalwart proponents.
ETHNews: Can you explain why the new family [of consensus algorithms] described by Team Rocket is a legitimate third family "type," academically-speaking?
Emin Gün Sirer: Indeed, we have seen only two main consensus protocol families emerge over the years.
The first one is "classical" consensus, used in permissioned blockchains and distributed ledger technologies (DLTs), the second one is Nakamoto consensus, also known as PoW.
The former is great in that it provides quick finality, but it requires knowing the precise set of participants, is fragile and does not scale well. The latter is great in that it is open to participation and no one has to know the full set of participants, but it's inherently slow, inherently not scalable, and is not sustainable.
This new family combines the best of both: very quick finality, without a precise notion of membership.
What makes it a different family is its mode of operation: it operates by polling a small sample of the crowd, and steering everyone towards the emerging consensus.
[By contrast,] classical consensus protocols check everyone, while PoW elects a leader, who determines block contents. This [novel] protocol family is very lightweight: it simply guides everyone towards wherever the majority is headed.
EGS: Some critics of Team Rocket's work have noted the synchronistic nature of their underlying protocol, essentially citing security concerns.
Generally, in computer science the highest standard of security – provable mathematically – for securing distributed systems is asynchronous byzantine fault tolerance (ABFT). While security is surely important, especially for systems where users do not know one another, it is becoming more and more apparent to some ecosystem thought leaders that security, as a feature, is part of a suite of functionality tradeoffs that must be reconciled to one another for certain milestones (like scaling the Ethereum blockchain) to be reached.
Despite Team Rocket's methodologies only satisfying the more rudimentary synchronous byzantine fault tolerance, the new family reaches a level of security that is simply good enough, while surging forward with other advancements.
Horseshoes, Hand Grenades, and Synchronous Blockchains
ETHNews: Can you explain how this new family is secured? Does it remove the golden standard of ABFT?
EGS: This protocol is proven safe for a synchronous network. The Team Rocket paper introduces a slew of revolutionary new proof techniques not seen in distributed systems work so far. So it makes sense that they would target a set of conditions that are easy to analyze.
But just because a protocol is proven safe under one model does not mean that it is safe under others. Team Rocket conjectures that their proofs will apply to partially synchronous networks, which, incidentally, is identical to bitcoin's safety guarantees.
Generally, with synchronous methodologies, there is a maximum amount of time between when a node sends a message and when that node can be certain the receiving nodes have received the message. This is verifiable mathematically.
Lab Conditions Vs. Real Life
EGS: While synchronicity has a certain elegance appreciated by some academic purists, asynchronous models are appreciated for more closely resembling real world scenarios, like the digital landscape of the internet, where nodes are constantly coming online and going offline and messages are dropped all the time.
ETHNews: If it's identical to what already exists [in bitcoin's safety guarantees], why do you think some have a problem with the lack of ABFT in Team Rocket's new family?
EGS: We often see academics insist on asynchronous protocols, which have the nice property that even if the Internet were to partition, they would not have a safety violation. But in my online lifetime, we never had a non-trivial Internet partition, so insisting on an asynchronous protocol seems misguided. And it seems clear to me that one could easily extend the protocols to detect partitions and take safe action.
It should be mentioned that Sirer acknowledged via Twitter that he and his team at Cornell had been asked to review the whitepaper previous to its publication on IPFS, and to provide feedback.
However, he made an effort to clarify that the innovations in the paper belonged to Team Rocket alone, and that he'd only provided peripheral guidance:
An Anonymous Avalanche
Emin Gün Sirer told ETHNews that he fully expects Team Rocket's underlying protocol to replace the consensus algorithms found in nearly every single transfer-of-value coin or token. "It's a great fit for [unspent transaction output] (UTXO) based coins, and if you have a PoW coin today, you should at least think about layering Avalanche on top, in addition to the security of the PoW layer, if not replace PoW entirely."
When asked why Team Rocket chooses to keep its identity a secret, Sirer said, "I suspect they chose to remain anonymous for the same reason most other people do. To disassociate the work from themselves, and allow the work to be judged on its own merits."