Thread by @gluk64

1/ Really well-written arguments by @kaiynne! A perfect starting point for the conversation about L2 scaling.

The crucial question not addressed in this post: what about negative network effects of OR (aka you can't securely hold >$50M in a single OR instance)?

A thread 👇

https://twitter.com/kaiynne/status/1318748044230160384

2/ As @VitalikButerin pointed out in his ETH Online talk, optimistic rollups expose stong anti-network effects: the more volume a single OR has, the less validators can/will run the full node, i.e. the less secure it is to rely 1-of-N honest assumption.

But it doesn't end here.

3/ Optimistic rollups can be hacked at L1 using a 51% soft-fork attack (with every asset stolen from them). Here is the description of this attack vector:

ethresear.ch/t/nearly-zero-…

4/ The nominal cost of such an attack is currently $256k/hour or $43M/week (see the link below). But the actual cost is much lower, since you only need a plausible commitment to force rational miners into compliance (see the attack description above).

crypto51.app/coins/ETH.html

5/ This attack vector cannot be mitigated as long as ETH mainnet remains PoW (quite some time yet).

Needless to say, these concerns do not apply to ZK rollups. Cryptographic assumptions aside, ZKRs offer exactly the same passive security guarantees as L1's.

6/ So, what are the implications of this? If a single optimistic rollup cannot hold substaintial value (in Defi we're talking about billions of USD), it has no way to become a single platform for composable Defi protocols. Period.

7/ So, optimistic rollups can't save Defi composability, and @kaiynne ruled out other solutions in his post. Is there really no hope? 😨

Well, let's dig deeper.

8/ "Even if ZKrollups progressed at a breakneck pace, all variations currently require rewriting contracts in a new language."

This used to be the case. But I'm excited to announce this statement is no longer true 😎

9/ It will very soon be possible to take any existing contract written in Solidity/Vyper and deploy it to @zksync with a minimum (if any) modifications.

zksync.curve.fi on our testnet had a hint at how this will work. Please stay tuned for a few more weeks!

10/ One last point from the post I'd like to expand:

"The final major concern is on the impact on cross-layer composability from withdrawal delays — a concern for basically all scaling solutions except alternate L1’s."

Correction: expect alternate L1’s *and ZK rollups*.

11/ 1-week withdrawal delays are indeed a hugh problem for optimistic rollups. Using payment channels to accelerate exist will be highly capital-inefficient (in other words, very expensive for users). Here is a great post about it:

medium.com/starkware/the-…

12/ But ZK rollups (such as @zksync) can provide finality in minutes and do not suffer from this problem at all.

This enables a gradual transition from L1 to ZK-powered L2. Here is how this will work:

13/ Users have their assets in a ZK rollup, enjoying L1-level security and scalable interaction with contracts in L2.

Whenever they need to interact with legacy protocols on L1, they simply do it with a few min delay. Moreover, many of these transactions (e.g. Uniswap trades)...

14/ ...can be bundled together and executed as one single tx on L1, thus effectively scaling L2<>L1 interaction.

15/ If this sounds too good to be true... well, it kind of is! ¯\_(ツ)_/¯

Zero-knowledge proofs are a once-in-a-generation breakthrough. Few understand. And we haven't even spoken about frontrunning-resistance—but that's a topic for another time.