Thread by @epolynya

Haven't listened to UCC, but seeing a lot of comments about it. Firstly, I'm delighted we have moved past the "rollups bad" phase to "rollups awesome but only on L1 bags I hold". However, the whole point of a modular architecture is the concept of L1s melt away.

1/

Rollups will always verify on the most secure, decentralized, robust, battle-tested settlement layer, backed by the soundest money and with the widest range and depth of liquidity. The cost is negligible, and the risks of having a bridge to a less secure SL isn't worth it.

2/

I believe Bitcoin would have been the best SL, but given it doesn't support rollups, Ethereum is the next best option. For a PLONK rollup doing 1,000 TPS, gas fees per transaction is only $0.00003. With 10,000 TPS it's 10x lower still etc. These are negligible numbers.

3/

So, to compete with Ethereum and Bitcoin, the only choice settlement layers have is to become more secure, decentralized, robust, and sounder money coordinating liquidity across more assets. Gas cost is an irrelevant argument as it is negligible, as mentioned above.

4/

Now, data availability is a more interesting space. Currently, Ethereum's data availability is certainly very expensive, so it makes sense for ZKRs to leverage alt-DA options. But, so far, both StarkNet & zkSync 2.0 are building their own solutions rather than using "L1s".

5/

Data sharding & data availability sampling invert the trilemma - so the more decentralized your network is, the more DA throughput there is. So, current L1s w/o DAS are totally obsolete anyway. Celestia & Polygon Avail are much better positioned to capture market share in DA.

6/

In short, rollups are inevitable. With settlement layers, you have to be the more decentralized, secure and robust - there's no other way. With DA layers, you have to implement DAS and also be most decentralized. Higher decentralization = higher TPS for rollups.

7/

I get it - the confusion comes from people still being entrenched with the monolithic blockchain mental block where decentralized and throughputs were trade-offs. This is no longer the case in a modular world - unless you're a rollup, you must compete on decentralization. Period.

Addendum: For ORs, using an expensive-to-run DA layer does not make sense given their honest minority assumptions, as @aliatiia_ has mentioned. I can definitely see lower value ORs on Celestia or Polygon Avail, but it makes very little sense on a high TPS "L1".

@aliatiia_ One last thing, because I wasn't clear enough. A volition like StarkNet or zkSync 2.0 will settle on the best settlement layer, and can have multiple options for DA where users & devs can make their own choice. So the idea of trading off is simply not there. Best of all worlds.

Clarification: a couple of yesterday's L1s are building data shards with data availability sampling, notably Ethereum & Tezos. I believe NEAR has plans too. But it's very likely new DA layers like Celestia, zkPorter & Polygon Avail will have a short-term time-to-market advantage.

Clarification required: these are the costs paid solely to SL to verify proofs & state transitions. DA & rollup costs seperate. The point here was - ZKRs would opt for validium/volition solutions (see below) instead of an insecure L1. Related:
polynya.medium.com/how-rollup-fee…

It's important to note that splitting DA and SL brings additional security assumptions, and rollups with shared DA and SL are far more secure. But the point is, it's still more secure than a rollup on a Iess secure L1, and volition gives user choice. Hope it makes sense now.

As for why, here's zkSync's take on the matter, and why they opted for zkPorter: zksync.io/zkevm/#what-is…

Of course, long term Ethereum is also building a highly scalable DA layer, so the need to split DA & settlement will be much lower for zkSync and others.