Key highlights from David Rosenthal’s speech on the blockchain
David Rosenthal is a distinguished computer scientist who was the #4 employee at Nvidia. He spoke at Stanford a few days ago on his views on decentralisation. I've always found it useful to read opposing views of something I believe in. It helps me refine my arguments and spot opportunities in the space.
Permissionless blockchains consume more energy
In a permissioned blockchain, all nodes perform the same calculation on the same data. In a permissionless blockchain, nodes can perform a different calculation on different data to reach consensus. Blockchains like Bitcoin and Ethereum are permissionless, and consume more electricity. This is because they are incentivised to use higher compute power and be the first node to get the calculation right.
Sybil attacks are critical in network design
In a Sybil attack, the attacker creates a majority by coordinating a set of seemingly independent participants. This is what many of us know as collusion. For collusion to succeed, the rewards from colluding must be greater than the cost of it. Sybil attacks are the biggest threat to permissionless blockchains.
Economies of scale lead to centralisation
As noted above, the only way to prevent a sybil attack is by ensuring that the cost of mounting one is less than the reward. To increase the cost of mounting an attack, participating in the network must be expensive. When participation in anything becomes expensive, economies of scale kick in, i.e. if I aggregate units it gets cheaper and participants benefit. But then, isn’t this just centralisation?
Proof of stake sucks
David argues that proof of stake isn’t great because its like a decentralised, centralised database. Meaning that its blocks of centralised data that are then decentralised. If it’s too small, it can be attacked easily. If it’s too big, nothing happens.
Centralisation and immutability pose risks
Mining on Bitcoin is controlled by no more than 5 miners. Mining on Ethereum is controlled by 2 miners. The degree of centralisation is a technical and legal risk. Frauds, scams and rug pulls in web3 have been and will be a big problem. Immutability makes this worse because, as we know, once a transaction is completed it cannot be reversed.
I’m a big believer in web3 but I enjoyed reading this piece. If you have the time, I suggest you do the same. I’m going to think through David’s points and respond to them once I’ve had a chance to process.
- Sybil attacks: prevent them by ensuring the cost to put one up exceeds the reward
- But then economies of scale is a fundamental problem for decentralised systems. Participation is expensive and therefore economies of scale help → centralisation.
- Proof-of-Space-and-Time attempts to make participation expensive by wasting storage instead of computation. What is this?
- Mining is highly concentrated:
- < 5 pools control the majority of mining pools in Bitcoin
- Similar for Ethereum
- Tether powers a lot of exchange (stablecoins) but lied about its reserves