Rollups are not tied to Ethereum. Rollup teams can and will deploy on robust security and/or DA layers. E.g. If Bitcoin were to add opcodes to verify STARKs, I bet we see StarkNet Bitcoin. Or if someone invents a revolutionary consensus mechanism that obsoletes PoS.
(1/4)
Like DeFi protocols, they may also deploy on more centralized and insecure security layers to preempt forks. I'd bet they would be incentivized to do so. E.g. Instead of $X00M on ephemeral liquidity mining, more pragmatic would be to pay StarkNet to deploy on Avalanche.
(2/4)
Of course, only Ethereum has an ambitious roadmap for rollups with global scale DA, and I'm not aware of any other project even attempting to build a security layer with robust economics and network effects. Could change in 5-10 yrs, of course. Niches can form elsewhere.
(3/4)
The reason I focus on rollups and not security/DA layers is because everything I've said about doesn't really matter long term. Like I argued in my last thread, the L1 fees will tend towards negligible - the real battleground long term will be between rollups themselves.
(4/4)
• • •
Missing some Tweet in this thread? You can try to
force a refresh
These numbers are measured, $0.06 tx fees is pretty much where I'd place a well optimized optimistic rollup. Just surprised they're already there given Arbitrum/Optimism are quite some ways away. Much easier with payment rollups and custom wallets, of course.
Looking at the big picture: $0.06 is with the highly constrained DA on Ethereum currently. Once modularized with data sharding, the floodgates will open. I suspect calldata will become negligible, and L1 verification and L2 fees will be the dominant costs. Previewed w/ validiums.
So, the cheapest rollups in the future would be:
a) With the highest activity, thus L1 verification cost for each transaction is amortized tending towards negligible
b) The most efficient rollups with the lowest L2 fees (sequencer & proving costs)
@guiltygyoza is building a universe on StarkNet. Now, monolithic chain (L1) apologists or rollup deniers may shout - this is a pointless, frivolous waste of time. The point is, if you can simulate reality - the universe - on StarkNet, you can build *everything*. (1/6)
For the longest time, I had thought it would be impossible to build a decentralized reality simulation - ala Existenz/RP1. But zkRs are that magical invention that can realize this. Eventually, you can build entire game engines and reality simulators within zkR clusters. (2/6)
All you need to decentralize anything is to generate succinct ZKPs and compressed outputs, and have them verified on a security layer like Ethereum. The real heavy lifting - simulating the universe - can be done on a zkR. Won't be long before we have Vulkan API on a zkR. (3/6)
dYdX already does complex DeFi derivative trades for $0.10 - effectively zero-gas abstracted from the user - and reducing more as activity ramps up. The first smart contract rollups - Arbitrum One & Optimistic Ethereum - are early MVPs with many missing features & optimizations.
With Nitro & OVM 2.0, both are seeing significant upgrades soon (OVM 2.0 in 2 weeks). As data compression systems come online, we'll gradually see ~10x further reduction in costs. Of course, StarkNet promises even lower costs - see dYdX as evidence - and alpha releases in Nov.
With the execution layer modularized, the final piece of the puzzle of modularizing data availability. Validiums & volitions are already doing this with <$0.01 fees possible today w/ scale. Rollups leveraging Ethereum will do the same when data sharding releases.
@KyleSamani@Evan_ss6@latetot Mentioned - "this is not just about Ethereum". Rollups maintain full composability across multiple DA sources - incl multiple data shards & even non-Ethereum DA (volitions). Long-term, one zkR can/will outscale Solana by up to ~10x with full composability and better UX, (1/4)
@KyleSamani@Evan_ss6@latetot And necessarily, zkRs have superior composability and liquidity (de)fragmentation properties, with initiatives like dAMM & Pooling, than between L1s or sharded networks (which Ethereum is no longer - they're only for data to feed rollups). It's better UX across the board. (2/4)
@KyleSamani@Evan_ss6@latetot The right comparison is StarkNet vs Solana, not Ethereum vs Solana. One StarkNet instance can outscale Solana, and you can have multiple StarkNet "shards" which can be very tightly coupled, sharing liquidity and other interoperability, and with SHARP share provers. (3/4)
Just showerthoughts as a non-expert on data availability for validiums: what if we could take StarkWare's adamantium concept and structure it into a permissionless DAC? Users can choose (& vote for) their preferred DA provider(s) or do it themselves.
Now, the voting may make it sound like DPoS, but you don't actually need to come to consensus. As long as an n number of DA providers (by vote) sign off, the validium can transition. If the validium freezes, you only need to trust your preferred DA provider or yourself.
Essentially, you have a baseline DAC that sign off state transitions - except instead of being permissioned it's voted in by governance. On top of that, you add a 1-of-N trust layer with adamantium. I could be wrong - but we're now approaching OR type trust assumptions.
This is the "cost of validator censorship", not decentralization. Validators merely provide a service to the network, i.e. ordering and signing transactions - it's non-validating full nodes that enforce consensus rules. Covered here: polynya.medium.com/security-layer…
- Ease of running full nodes. If users must trust a validator set, then it's not a trustless network - you're just another bank with a new set of bankers, as Gavin Wood put it.
- A wide token distribution with high monetary premium to increase difficulty of validator censorship.
I'd also add that Solana and monolithic chains are highly inefficient, requiring thousands of nodes for minimal security. Long term, validiums will be a fraction of the cost. A succinct ZKP can verify millions of transactions, while DA layers can have 1-of-N security models.