In a recent Bankless podcast, they had on Anatoly Yakovenko and Emin Gün Sirer of Solana and Avalanche, respectively. Unfortunately, the interview was pretty light on technical details. That said, there was one interesting part where they discuss philosophy on L1 vs. L2s and both Anatoly and Gün suggested that the presence of an L2 was basically a failure on the part of the L1.
As a reminder, Layer 1 refers to the consensus layer: You have transactions that are happening and need to be properly ordered. Layer 2 theoretically sits on top of Layer 1, batches, and then writes down to Layer 1.
Anatoly basically says that the only reason to do L2 is if L1 is not fast enough:
Defining L2s to strictly this idea of something like optimism or Zero Knowledge Proofs that verify that some computation was done correctly, those are implementation details. Anybody could build this on Solana so they question is why would they want to, right? What is it that they're trying to do with this thing and the only reason I'd want to do this is because the execution layer and Solana is not sufficient for them...that it takes too long for some really big computation. But right now, that's not a bottleneck and it doesn't seem like that's going to be a bottleneck for the next couple years...we'll see, I don't see it as a necessity. – Anatoly Yakovenko
Gün believes that anything that can be done on L1 should be done on L1 since the batching that happens at the L2 level introduces the potential for nefarious behavior:
Every Layer 2 solution also works on Avalanche so we can accommodate anything Layer 2 that works on Etherum just as well but that's not where our scalability roadmap is taking us. I understand where Ethereum wants to go with its own Layer 2 based solution and I'm not headed in that way. I'm not headed that way because I have some concerns about where one might end up if one were to go that way and that is centralization land. All of these Layer Twos that exist today have architectures [that] lead to constructs where some central party is collecting all the rent...So you can end up wanting to build a new system that's really really scalable and you end up reinventing the system as we have now. Congratulations, you have Robinhood and Citadel and I don't want to end up there. So that's why we have so much emphasis on Layer 1 scalability. Everything that can be done on Layer 1 should be done on Layer 1. These L2s essentially take an opaque blob, do something to it, and bring it back into the system. That process happens away from users and there are many different paths for malfeasance where people can reorder what happens at the Layer 2 level so that's not where we're headed. – Emin Gün Sirer
It's not surprising that the the head of Solana and Avalanche, two of the most performant blockchains in the world, would extol the virtues of as much computation being pushed to the Layer 1 as possible. In a future post, I'll have to discuss how this vision differs from that of Stacks (STX) which is itself an L1 but uses BTC as a settlement layer.