The update this month had some very important additions including a new blockchain developer, improvements to Agoras Live both UX and wallet integration, improved Tau IDE, and a path to main-net. I will discuss some of the points in the video below and provide a running commentary.
The MARCH 2021 UPDATE:
Verifiable computation
Some of the highlights of this update which caught my attention was the addition of Lukas Saul and his discussion of a new algorithm to certify the correctness of a computation. This Tau computation correctness certification mechanism immediately directed my attention to Truebit. Truebit solves a very similar problem from a different approach which I think is synergistic with the cryptographically certified correctness approach. Truebit uses the game theoretic mechanism design method of answering the same question which is: "How can we know that a computation is exactly as promised across an unlimited number of nodes?". The Tau approach is to guarantee the correctness at a different layer where Tau gets a cryptographically certified versioning system which uses proofs (for proven computation).
The lowest layer is the human layer where mechanism design and game theory plays a role. How efficient is the game theoretic mechanism design method vs the cryptographic method? Both of these methods fit under the definition of verifiable computation. The important of verifiable computation is for both security and scalability of Agoras.
Probabilistically checkable proofs are a likely method Tau can use to achieve verifiable computation. Truebit on the other hand relies on a "“verification game" which reaches the same conclusion. I am curious to know which method is faster in which circumstances and how to combine these methods to achieve a high level of scalability along with the verified computation.
Note, the Monoxide network, Harmony One, and Zilliqa, rely on sharding which is scaling at Layer 1. Layer 2 scaling would be rollups, zero knowledge proofs, and outsourced computation. With regard to verification, should this be done on chain or offchain? At Layer 1 or on Layer 2? Right now I do not know how efficient Tau can be but my guess is to scale it will require a great deal of outsourced computations which means we need a high degree of correctness verifiability.
So far I think with this verifiable computation component we will know that the computation is verifiable across an unlimited number of nodes as long as we have the cryptographic proof. This is in my opinion essential and leads me to some further questions. For example what data structure will be used to house the computation? For example will we see a DHT (distributed hash table)? Will this be proof of stake and if it is proof of stake what algorithm? Considering that computation can be verified what option does that give with regard to proof of stake?
References
- https://truebit.io/
- https://www.pepper-project.org/
- https://www.semanticscholar.org/paper/Verifying-computations-without-reexecuting-them%3A-to-Walfish-Blumberg/52a8afc78ef5e312d0e90bbb6d92fdce7db6df1c
Zhou, Q., Huang, H., Zheng, Z., & Bian, J. (2020). Solutions to scalability of blockchain: A survey. IEEE Access, 8, 16440-16455.
Thank you for the references!