이더리움 클라이언트: 다음은 무엇인가?

Ethereum Korea One: Genesis · Day 2 · 2026-09-29 · 32:44

What comes after Glamsterdam for the software that runs Ethereum? At Ethereum Korea One: Genesis, client developers and a consensus researcher discuss ePBS, L1 scaling, ZK proving, and what Ethereum clients may look like in five to ten years.

YouTube에서 보기

연사

Jun Song, Core Engineer, Offchain (Prysm)
Felix Lange, Geth Lead, Ethereum Foundation
Luca Zanolini, Researcher, Alumni of EF
Youngseok Yang, CEO, Nodeinfra (모더레이터)

세션 소개

Moderated by Youngseok Yang, CEO of Nodeinfra, this panel brings together Jun Song (Prysm), Felix Lange (Geth), and Luca Zanolini, formerly of the Ethereum Foundation's protocol consensus team. They cover what ePBS changes for consensus clients and the builder market, the shift toward scaling L1, whether ZK and PeerDAS really resolve the trilemma, and the long-term future of clients and nodes.

핵심 요약

Jun Song said ePBS (EIP-7732) adds complexity: validators take on a new duty in the Payload Timeliness Committee (PTC), and fork choice must now handle three cases per slot instead of two (a missed slot, a block with its payload, and a block without it).
Song said that with today's attestation deadline four seconds into a slot, there are only a couple of seconds to execute and validate a block, and ePBS expands that window by up to five times. The target is to raise the gas limit from 60 million to 200 million, which he called conservative since the team wanted to ship 300 million at first. Auctions that happen opaquely in relays today move into the protocol, making the bid floor visible and letting even solo stakers bid.
Luca Zanolini said that bringing builders into the protocol makes it possible to prove which properties hold under which assumptions, for example that the protocol can force a builder to pay the proposer even if it withholds the payload. The trade-off is more in-protocol participants and more assumptions to sort out.
On L1 vs. L2, Song called the rollup-centric roadmap a success and expects L2s to shift toward specialized blockspace for institutions, since gas on L1 is now cheap enough for most apps. Felix Lange said "we always wanted to scale L1" and does not see it as exclusive with the L2-centric roadmap. Zanolini advised builders to follow Ethereum's principles (maximize trustlessness and decentralization for a neutral settlement layer) rather than specific roadmap entries.
On the trilemma, Zanolini said ZK and PeerDAS move assumptions rather than remove them (enough custodians hold data, at least one live prover exists), so proving must stay cheap to avoid prover centralization. Lange was disappointed that zkVMs rely on special system calls for operations like SHA-3, calling it "a bit like cheating." Song supports Optional Execution Proofs (EIP-8025) but noted PeerDAS still relies on supernodes, while Lange described PeerDAS as a stepping stone toward full DAS.
Looking ahead, Song expects at least five hard forks by 2030, with post-quantum (he prefers an Offchain Labs approach that reuses BLS) and formal verification in Lean as big agenda items. Lange said nodes will remain as long as someone needs Ethereum state, with a next version of Geth's Snap sync built on block access lists (BALs, EIP-7928). Zanolini sees clients moving from "re-execute everything" toward "verify everything."

챕터

← 영상 목록 · EK1 Genesis 행사 기록