← All calls
ARBORIST#124

Talks on speeding up ZcashD deprecation

As AI-driven security findings continued, options for bringing forward the ZcashD deprecation schedule were discussed.

Agenda summary 10

  1. Zebra update

    Zcash Foundation · Marek

    • Over the past two weeks the team handled the reported security advisories.
    • Right after the previous call it released Zebra 4.3.1 with some of the fixes, and will soon release a new version with additional fixes.
  2. ZODL update

    ZODL · Daira Emma

    • The team released sapling-crypto 0.7.0, orchard 0.13.0 and the entire librustzcash stack.
    • The old core2 crate was yanked by its original maintainer, so the team built a replacement crate itself, which does not depend on unsafe code and also supports no_std environments.
    • Zallet wallet import can now be regression-tested with history going back to genesis.
    • A regtest wallet for testing was created and verified, and a PR is up in the integration test repository.
    • On the research side, it merged a cleanup PR for the quantum recoverability specification and is working on turning the earlier informal security argument into a complete formal proof.
    • It also tidied up a repository of AI agent skills for working on the Zcash codebase.
    • It shipped zcashd 6.12.1, a security release, and there is a post on the ZODL blog.
  3. Zaino update

    Zingo Labs · Hazel, Alfredo

    • There were few publicly visible changes and a lot of internal work.
    • The team sped up tests and improved integration tests, and fixed APIs that had diverged from lightwalletd (mainly version numbers).
    • It upgraded to a new dependency on core-zed instead of core2, improved graceful shutdown and cleaned up duplicated code.
    • Alfredo flagged a dependency issue zancas filed on Zebra that causes problems for Zaino, and Hazel said it was new to them and got the link.
  4. zcashd deprecation update

    Electric Coin Company, Zcash Foundation, Zingo Labs · Pacu

    • A new Zaino release is being prepared, and the team is waiting for release manager Nacho, who is traveling.
    • Zaino was bumped to Zebra 4.3.1, and untangling the dependencies made Zebra version updates much faster.
    • It sorted out transparent-data-related problems in the Zaino-lightwalletd parity integration tests, and the PR is awaiting review.
    • The light client working group meets every other Thursday at 17:00 (UTC).
    • The block explorer RPCs are mostly done, and for the few remaining, the team will check whether they are actually used and then wrap them up.
  5. Accelerating zcashd deprecation and discussing security vulnerabilities

    Daira Emma, Arya, Zooko, Pili, Pacu, Jason

    • Daira Emma argued that AI vulnerability discovery has become serious and that zcashd must be deprecated sooner.
    • 6.12.0 fixed a bug where Sprout proofs were not verified in some cases.
    • 6.12.1 fixed a crash bug in zcashd and Zebra related to the identity point, a consensus divergence between the two implementations, and a bug that bypassed checks.
    • In response to the possibility of undefined behavior in balance arithmetic, it also added a checkpoint mechanism that restores balances stored on disk.
    • Daira Emma proposed releasing security updates on a predictable date, like Microsoft's Patch Tuesday, and simplifying the protocol to remove the Sprout code.
    • Arya recommended putting zcashd behind a Zebra node, and Daira Emma countered that this exposes it to both Zebra's crash bug and the consensus divergence, so it does not get only the advantages.
    • Zooko said that Shielded Labs keeps recommending this to partners who cannot turn off zcashd right away, despite knowing the risk of a chain fork.
    • Zooko said that Shielded Labs is hiring Taylor Hornby to audit Zebra and zcashd.
    • Zcash Community Grants announced a bug bounty totaling 1 million dollars, with a maximum of 225 thousand dollars per report.
    • Pili mentioned that duplicate reports are increasing, so low-severity issues may already have been found.
    • Pacu explained that, in the direction Kris proposed, the plan is to give up the frictionless transition, functionally match the most important use cases, then greatly strengthen outreach and support to bring deprecation forward.
    • Jason said that since ZSA is not included in NU 7, the next network upgrade should not be blocked because of zcashd deprecation, and that if deprecation is hard within a few months, the upgrade should be done in both zcashd and Zebra as a candidate with an agreed scope.
    • Daira Emma said not to keep changing the NU 7 composition that has already been decided.
  6. ZSA update

    Qedit · Vivek

    • An end-to-end demo of the full feature set, covering issuance, transfer and atomic swaps, was held at ZecHub and is on YouTube.
    • It can be run locally or tried by connecting to a Qedit node that supports swaps, and a forum post is coming soon.
    • Since ZSA is not included in NU 7, the team started discussing merging the swap's consensus content into the ZSA ZIP instead of leaving it scattered across ZIP 226, 227 and 228.
    • The Orchard PR has answered all of Str4d's review comments and is awaiting a response on merging into the long-lived ZSA branch.
    • The structural changes for swaps will be discussed with Daira Emma and incorporated after the review is finished.
    • Other crates are also ready for review, and the last PR for Zebra is also ready.
  7. Network Sustainability Mechanism (NSM) update

    Shielded Labs · Judah

    • Conrado found a small mistake where a zero value was used for the deferred pool balance change.
    • The parameter was passed so that the procedure computes the correct value, all call sites were updated, and a regression test that catches such mistakes was also added.
    • Once this PR is merged, the later PRs can be reviewed, and additional review is needed.
  8. Crosslink update (Trailing Finality Layer)

    Shielded Labs, ZODL · Mark, Pacu, Daira Emma, Zooko

    • The feature net opened on April 16 and is supporting participants.
    • Six people made 91 commits, and at least three community block explorers were built, including a Crosslink version of Cipherscan.
    • The first PR from the external community was merged.
    • The core work is a custom UDP transport layer for block sync (codename shielded transport pigeon).
    • It attaches IDs to UDP datagrams, encrypts with Noise using the snow library, and reassembles up to 8MB.
    • Noise has no post-quantum support, so adopting clatter is being considered, and for now migration debt is accepted.
    • Block sync was replaced with this transport layer, and the mempool uses the existing approach.
    • It strengthened real fee values and staking window verification, and T-address and unified address wallets and funding streams work.
    • It is building the scanner and coinbase for incentive rewards, and next are staking rewards, the finalizer commission, mempool sync and a review of flyclient height proofs.
    • Daira Emma criticized building a new UDP transport layer when proven protocols like QUIC exist, saying it is a large maintenance burden and goes beyond Crosslink's original scope.
    • Zooko answered that as chief product officer it is their decision, and that the plan is to use this transport layer in season 1, improve it for use in season 2, and then reconsider.
  9. Dynamic fees update

    Shielded Labs · Mark

    • There is no big update.
    • There were personal circumstances, and a blog post about recent work is coming soon.
    • The speaker said there would be good news in a day or two.
  10. Open discussion: formal verification of Rust code

    Natalie, Daira Emma, Conrado, Pacu

    • Natalie asked about the status of formal verification of Zcash and Zebra's Rust code.
    • Daira Emma enthusiastically supported the grant proposal to verify Halo 2 in Lean 4, and said that if possible they want to verify the whole protocol in Lean 4.
    • They said AI has become very capable at formalization and Lean 4, making possible what was hard six months ago.
    • However, they said that verifying designs is more realistic right now than translating Rust code.
    • The parts to verify are those with inherent complexity, such as elliptic curve optimizations and the Sapling and Orchard circuits, and they said zcashd's C++ code cannot be verified and should be deprecated.
    • Natalie mentioned an elliptic curve verification project that translated Rust into Lean with Aeneas, and Daira Emma answered that they know of it but it is not their main interest.
    • Conrado confirmed that the Zcash Foundation is not currently doing formal verification of Zebra.
    • Pacu asked people to comment, since the Halo 2 formal verification grant is still receiving feedback.

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