I’m the founder of this project that just came out of stealth. Some more info in this article. Our Twitter if you want to follow us.
I’ve been in the crypto space for years and am an ICP seed investor, so well acquainted with the IC.
My current very tentative preference is to build this application on the IC and an Ethereum layer 2 (most likely Starknet as it looks like it’ll be the cheapest Ethereum L2 and we’re an extremely fee sensitive application - we expect micropayment tips to be the main source of protocol revenue).
Since I expect to be asked why, the rationale for using two chains is fourfold: 1. We need to decentralise the entire stack so as not be deemed a publisher and there are few blockchain-type systems that can handle something like this, hence the need for something like the IC. 2. However, the IC isn’t particularly well-suited to DeFi and our protocol, from funding to quality control is entirely market-based so we think we need another chain for that element of our protocol. 3. Even if it could do DeFi well, the IC, in my view, will never be as resistant to coercion as Ethereum POS is (secured by pseudonymous nodes) so I would prefer the money part of our app to be primarily secured by Ethereum. I’m not particularly worried about IC nodes colluding to steal, but I would be worried about governments successfully pressuring them to seize the assets of journalists that criticise them. IC nodes can be easily located.
Our goal to use IC over Starknet naturally requires one of the great innovations of the IC: chain key crypto (And yes I know if you use chain key you’re reliant on the IC’s security but only on a per TX basis unless I’m mistaken. I’m OK with that situation). However, currently only Bitcoin and Ethereum are supported and if I’m not mistaken Solana is next in line? So my question is, what plans, if any, are there to extend this functionality to Ethereum layer 2s and specifically Starknet? I’m guessing the prioritisation is done by usage and since Starknet currently has very little usage it might be a long way down the list, if it’s on it at all? FYI, we’re probably two years away from launch, this is a really complex protocol we’re attempting to build, so no major rush on this, but naturally it’s still something we’d need to know before we commence development in earnest.
If anyone involved with Dfinity has an answer to this question, or can point me in the direction of who to talk to, I would appreciate it.
I think this makes perfect sense and I expect this to be the primary architecture for web 3 dapps: Bitcoin for payments, traditional smart contract chains like ETH/SOL for DeFi and simple DAO functionality, and IC for everything else.
I wouldn’t have expected Starknet to be the most cost-efficient L2. Do you have a source for that?
If Starknet is your choice, then you could still interact with it given the current infrastructure ICP provides:
Starkent has built-in account abstraction, i.e. you can have an account that is controlled by a tECDSA or tSchnorr key controlled by a canister. I assume that there are Cairo implementations of this already, but I haven’t checked. Hence, I don’t think there’s any protocol-level work on ICP needed to interact with Starknet.
In order to communicate with Starknet, you can check which of the public RPC services support IPv6 and use HTTPS outcalls to call them directly. Creating an RPC Canister similar to the EVM RPC Canister that supports Starknet would also be a manageable effort and wouldn’t necessarily be done by DFINITY.
Thanks! Took much longer than expected to come out of stealth but the extra R&D was worth it. Our mechanisms are pretty tight now. Itching to build the thing at this point.
OK yes of course Starknet has native AA and that makes integration pretty straightforward. Excellent.
Re costs, currently Starknet is quite competitive according to this list. However I expect it to be even more so in the future as given the trajectory of computation costs, ZK proofs should continue to improve relative to Optimistic Rollups. Relative to other ZK Rollups they’re already very competitive as much like ICP, the didn’t just reinvent the EVM wheel but built everything from scratch to optimise for the problem they’re trying to solve. Eg their proofs are much cheaper than the competition. They’re also introducing TX parallelisation this year, which should be another large efficiency gain. So my position is as much a bet on the future rather than a reflection of present costs.
Would love to hear your or anyone else’s views on this though. We’re only beginning this process now after spending so long focusing on our own protocol design.