- There's a lot of aspiration, but the technology is far behind.
- Supposedly, you can only get 2 out of 3: security, scalability, and decentralization.
- Not acceptable, and not true.
- Every blockchain has two parts:
- Build a chain (we know how to do this, many good data structures)
- Choose the next block
- PoW
- Very expensive, unscalable
- Very centralized (BTC mining controlled by 3 mining, 2 of which owned by Bitmain)
- Delegated PoS
- "Look at these 21 people in charge of making our blockchain"
- Not very decentralized.
- Bonded PoS
- The amount of money you bond determines how much control you have.
- If you behave badly, you are punished.
- Does this work?
- Well, most people can't participate in this system. Most people have no disposable income to stake on blockchain behavior.
- Plus if blockchains support the economic system of the world, there's too much value at stake relative to the amount of bonding.
- No punishments needed—cheating is not possible.
- Money is always at your fingertips: nobody needs to bond anything.
- If most of the money is in honest hands, then the system works.
- More specifically:
- Each token has the same decision power
- Uses Byzantine Agreement to achieve consensus
- Old Byzantine Agreement protocols could seldom handle more than ~15 parties
- Otherwise very slow, and have a fixed numbers of players
- Algorand works in 2 phases
- A random user is selected among all global users (randomly sampled according to coin distribution)
- The user proposes, signs, and publishes a block
- 1000 users randomly sampled among all users, their keys are known to all users, and they validate / reach agreement on / sign the block proposed by the first user
- If 10% of users are bad actor, probability that majority of 1000-person committee is Byzantine is negligibly small
- This subset of users can run Byzantine Agreement to achieve consensus
- Committee "selects itself" using verifiable random function (VRF)
- Think of it like a lottery you run in your own machine—if you win, you propagate that winning ticket to the rest of the network (and can be universally acknowledged)
- This is fast because this VRF can be evaluated within a microsecond
- Committee is not manipulable, even by a powerful adversary
- Agreement on the fly while the block propagates, as opposed to separating into two phases
- And single round of voting finalizes the block
- This makes it super fast
- However, it's also resilient to network partitions
- If powerful adversary tries to hold back messages and secretly finalize a block, that block will also be publicly finalized
- Smart contracts
- Secure incentives
- Dutch auctions
- Algorithmic stabilization
- Minimized memory
- Random-access chain
- Private chains
- Treasury bonds
- Secure self-governance
- Named assets
- Privacy & correctness