Light Client Task Force #2 - Multiproof Formats & Proof Negotiation

[wemeetagain]

DATE: Dec. 4, 15:00 UTC
DURATION: 1.5 hours
LOCATION: Zoom

Purpose

Discuss upcoming light client research
Coordinate to bring ethereum light client tech to production

Audience

Researchers, client developers, wallet developers

Agenda

1. Introductions (10min)
2. Discussion on multiproof formats and proof negotiation (eth2-focused) (50min)
3. Next Steps / Tactics / Open Discussion (optional) (30min)

Proofs are the lifeblood of any light client. We will discuss the landscape of multiproofs. Proof backings / concerns / consumers / request/response interactions / interfaces.

Background Reading

• Current SSZ Spec - describes the mechanics of SSZ merkleization
• SSZ Multiproofs Spec - describes generalized indices and multiproofs
• New SSZ Spec - WIP, includes more motivation and detail around the merkleization/proof side of SSZ
• Summary of SSZ Multiproof concepts / backings - see "RE multi-proof structure:"
• SSZ Sparse Merkle Tree (SMT) - discussion about adding SMTs to SSZ
• EE Multiproof Concerns - great writeup on the expected EE toolchain
• Experiments w Merkle Trees/Tree-offset Proof Backing
• Discussion on Requesting Multiproofs
• Discussion on SSZ Partials / Classic Proof Backing
• Turboproof Proof Backing

Discussion Ideas

• How can we create a unified infrastructure for requesting / handling proofs across phase 0/1 data and EE data?
• How should proofs can be requested/responded by light clients and servers? Where is the overlap between light clients handling Ph0/1 data proofs and block producers handling EE state proofs?
• [your idea here]

Previous Meetings

• Meeting #1(https://medium.com/chainsafe-systems/light-client-task-force-call-1-1aaf559230fb)

[zmanian]

Light client proof formats are critical for IBC.
Here is some related work.

https://github.com/confio/ics23

[protolambda]

For anyone who is interested in implementing partials, or any kind of specialized data-sharing ssz structure (great for caching hash-tree-root!), I made a write-up about implementing it, with some early results: https://github.com/protolambda/eth-merkle-trees/blob/master/typing_partials.md

[wemeetagain]

Recording: https://www.youtube.com/watch?v=h_GeWO0U-bw
Transcript: https://medium.com/chainsafe-systems/light-client-task-force-call-2-bdb3d24d2b2b