Paczki @paczkiii - Twitter Profile

Each Solana transaction (demo for the "web-scale blockchain") is capped at 1,232 bytes, because the whole thing, signatures included, has to fit in a single UDP packet (IPv6 MTU is 1,280 bytes, minus 40 for the IP header and 8 for the fragment header; they sized a global exchange to a datagram, but we'll ignore that). Nasdaq pushes ~20 billion messages a day, about 231,000 a second; assuming mainnet comes online in 2020 and hits real Nasdaq volume by 2025 (Anatoly's own pitch is "Nasdaq at the speed of light"), you'd need 3.5x the claimed 65,000 TPS, but we'll be generous and use their number. That's 80 MB/s of raw ledger, 6.9 TB per DAY. The official validator requirement is supposed to be "commodity hardware." Solana has one facility for block production at a time, the "leader" (it rotates maybe, but a leader in Singapore means votes from Frankfurt would GRAZE the 400ms slot boundary- I have a feeling finality won't be happy about that...) So the entire validator set would have to sit in the same building. We already invented that. It's called us-east-1. And say the chain somehow works. Where's the business? Wallets? Wallets are free open-source commodities. The Fat Protocol Thesis (2016, settled science) proves 100% of value accrues to the L1; a wallet that charged a fee on a swap would lose every user in one epoch. The total addressable market of humans willing to hand-copy 24 words onto paper is maybe 40,000 people, and they all already own a Ledger. Also, an on-chain order book means a single market maker quoting 200 pairs eats 3% of the entire chain's capacity by himself. Order books run on servers. This is settled. Let's be VERY optimistic. Every validator runs an expendable EPYC with 256 GB of RAM, every transaction is EXACTLY 1,232 bytes (unlikely, you need margin, and likely priority fees from having a congested network), we'll ignore whether the state can actually fit in RAM, we'll ignore MEV entirely (landing in the wrong slot means hell for an arber wanting to exit; meaning 160 validators per metro (almost certainly impossible due to not enough dark fiber)). That means doubling effective throughput every year, minimum (almost certainly unrealistic.) We'll also assume the network never halts once, starting Jan. 1, 2020, and give it until Dec. 30, 2027 to hit Nasdaq volume. Sounds easy, right? But that's ~23% faster than Moore's Law has ever moved, every year. The ledger would literally double underneath the hardware meant to store it, at the absolute bare minimum. A minimum which is almost certainly physically impossible. I don't think shipping the software will be the issue, Anatoly... No one at Solana Labs did these back-of-the-envelope calculations, and asked Anatoly- maybe it was a little ambitious?

15

81

5

15

12K

0

16