← All calls
ARBORIST#130

Debate over the fee parameter change process

A procedural debate broke out over reflecting fee parameters in implementations before a ZIP existed.

Agenda summary 8

  1. Zebra update

    Zcash Foundation · Alfredo

    • On August 10 the team released Zebra 6.3.0, which includes security fixes related to syncing and peer scoring.
    • It now implements a compact transaction streaming interface through a lightwallet-compatible gRPC server, so light clients can connect directly to Zebra without lightwalletd.
    • It is experimental and off by default.
    • On the consensus side, the transaction verifier was split into a block verifier and a mempool verifier.
    • Banning a peer now removes all address book entries for that IP, so a malicious peer cannot return to the address book.
    • The peer set was simplified and the address book was moved behind a Tower service.
    • An external contributor added fuzzing coverage to the network and consensus crates, and it is under review.
    • The team switched to the Changie tool to reduce changelog conflicts.
    • A large refactoring PR that replaces the transaction type with the one from zcash_primitives will be ready soon.
    • Flaky CI tests are being moved to nextest, and Arya and the ZIP editors are working on five draft PRs for a new QUIC-based P2P protocol.
  2. ZODL: core libraries and Zallet update

    ZODL · Pacu, Alfredo

    • Most of the findings from the Zallet security audit were resolved, and gRPC and other parts were hardened.
    • The zcashd migration path was also hardened with feedback from miners and users, so it can import pre-Sapling wallets without an HD seed and old key material.
    • Crates such as zcash_client_sqlite 0.22 were released, and this crate release is final, with no more RCs.
    • RPCs such as a sign-message RPC were also added to Zallet.
    • The Ironwood pool migration engine also progressed.
    • Pacu asked that whoever made improvements to the Orchard crate and librustzcash upstream them, saying Pacu would review and merge them.
    • Alfredo said they would check the Zebra regression issue Pacu had filed that day.
  3. ZSA update

    Qedit · Vivek, Daira Emma, Alfredo

    • The main work was catching up with the NU 6.2 and 6.3 changes.
    • NU 6.2 was applied across Orchard, librustzcash, Zebra and the transaction tool, NU 6.3 was applied to Orchard first, and the rest is working on the Ironwood changes.
    • After discussion with Daira Emma, it was decided to split Orchard's circuit into a separate PR.
    • The Halo 2 PR needed for the Orchard ZSA circuit has finished review and may be merged or included in the next release.
    • At the request of the ZIP editors, a draft swap ZIP was added to the upstream zips repository as ZIP 228.
    • With both swaps and ZSA getting ready, the team is examining combining them into one release and the workload for each crate.
    • Daira Emma said the Ironwood formalization is nearly done and that over the next few weeks they would focus on reviewing these PRs.
    • Alfredo said the refactoring that replaces Zebra's transaction type with the one from zcash_primitives may increase the changes in the Qedit PR, but that the code has changed greatly anyway because of Ironwood.
  4. Network Sustainability Mechanism (NSM) update

    Shielded Labs · Judah, Daira Emma

    • As requested, the team is submitting a ZIP 234 draft that keeps halvings.
    • The aim is for the proposal to be backed by a ZIP at the very least.
    • Daira Emma suggested presenting the flattening option and the halving-preserving option as parallel tracks, or leaving one as another draft ZIP, so they can be compared in rendered versions.
    • This is because people will have to vote on these options.
  5. Crosslink update (Trailing Finality Layer)

    Shielded Labs · Philip, Daira Emma, Nate

    • At the August 26 workshop at 16:00 (UTC), the team will share stability improvements, the hard fork tool, the second payout date, plans for the next network implementation, plans for regular network resets and the new nightly network, and hold a Q&A.
    • It released Crosslink node versions 12 and 13.
    • Version 12 improved the reliability of staking and unstaking as the chain gets longer, thanks to a speedup of the built-in wallet, and also improved the sidechain visualization and the headless user experience.
    • Version 13 is a wallet stability patch.
    • No more systematic or widespread forks have been seen since v11, so that investigation was closed.
    • The first network restart is being prepared and includes the consensus changes learned from the first network and merging Zebra updates such as Ironwood.
    • When Daira Emma asked about the consensus changes, Philip cited a change to the staking data format.
    • A from field indicating which finalizer a redelegation moves from is added to make it easy for the hard fork tool to check the slashing window, and unimplemented features such as the 10% finalizer commission will also be added.
    • Nate added that backports of mainnet changes are also a category.
  6. Dynamic fees and the specification process discussion

    Shielded Labs, ZODL · Mark, Daira Emma, Alfredo, Zooko, Nate

    • Mark posted a PR with a 2000-series ZIP draft that lowers the marginal fee from 5000 zat to 1000 zat.
    • Also, Dave of Zakura raised the weight ratio cap tenfold, and after talking with Arya in DC, Mark opened a PR for the same change in Zebra.
    • Daira Emma said that changing an implementation unilaterally without changing the specification first makes it impossible to keep the specification and implementation in sync, and in particular objected that this change was never mentioned in the ZIP sync meeting or elsewhere.
    • They also pointed out that there was no public discussion of interoperability when 4x and 10x nodes are mixed.
    • Mark said they see little harm because it is an inclusive change that broadens the limit, but acknowledged there was no public discussion and said they would write a separate ZIP.
    • Zooko said fees can be left to wallets and miners to decide, and noted there are also opposing views.
    • Nate proposed as a new norm the minimal courtesy of informing ZIP authors in a non-blocking way when implementing differently from the specification, and said mainnet experiments are also worthwhile if judged safe.
    • Daira Emma countered that to update the security analysis the protocol must not move too much, and that if every implementation fails to follow the specification, workaround code and security bugs pile up as with internet protocols.
    • They also said that if ZSA had been handled this way, the coinholder vote would have been ignored, and urged caution.
    • Zooko said they had earlier said that in the early days of dynamic fees a ZIP and a consensus process were needed, then admitted they were wrong, explaining that they changed direction after judging over dinner that only wallets need to be persuaded.
    • Mark said they do not want dynamic fees to become a battleground for such disputes.
  7. Formal verification progress and Lean study

    ZODL · Daira Emma, Nate

    • Daira Emma introduced the group hash indifferentiability page in the Ironwood book (github.com/zcash/Ironwood).
    • The unknown-discrete-log property of the hash that produces curve points is needed for the security of the high-level protocol and Halo 2, and was a part missing from the formalization.
    • The page is backed by Lean code and states what is and is not proved and what must be trusted by relying on mathematical papers.
    • Going forward, they want to bring the other parts of the formalization up to the same level of explanation so they can be verified without reading hundreds of thousands of lines of Lean code.
    • Nate said that some Zcash participants want to learn Lean and proposed reviving the learning study club.
    • Daira Emma said it should cover both how to understand without AI and how to keep AI from getting ahead.
  8. Grant presentation: constant-time hardening of Zcash cryptography

    Erys · Facundo, Daira Emma, Alfredo

    • Facundo of Erys introduced a grant they proposed while contributing to librustzcash.
    • The goal is to use a threat model to find the paths in the Zcash Rust repositories that handle sensitive data, specify with code comments and tests where constant-time operations are needed, and also mark where variable time is acceptable because an adversary cannot reach it.
    • The support needed is little: defining and validating the threat model, answering occasional questions, and reviewing short PRs, so it does not take much time.
    • Daira Emma said it is useful and that they would make time.
    • They said that the multiscalar multiplication in Halo 2 proofs handles secret witnesses but uses variable time for performance, which may be fine thanks to blinding, and that such conscious choices should be documented.
    • When Alfredo asked why librustzcash was chosen, Facundo answered that, being new to the ecosystem, they started from a widely used repository.
    • Daira Emma advised that the actual cryptographic code lives in dependency crates such as pasta_curves and orchard, so these should be included in the scope.

This summary is based on the original meeting notes; check them for exact details.