Major overhaul of Zebra's transaction type
Zebra merged a transaction type refactor of more than 5,000 lines, with fixes still needed before the next release.
Agenda summary 7
Zebra update
Zcash Foundation · Alfredo
- The biggest change is the refactor of zebra-chain's transaction type.
- Zebra's transaction became a wrapper around the upstream zcash_primitives transaction type, touching more than 5,000 lines and removing a roadblock for other integration work.
- After the merge, Conrado found and fixed a performance regression in which script verification loaded transaction inputs a quadratic number of times.
- A fix for a consensus divergence reported by an external contributor is also in progress and is blocking the next release, so the schedule is undecided.
- The team said it will make reviews stricter.
- On the network side, the peer connection limit was tightened to apply per IPv6 group rather than per address, and a peer mempool download slot that was not freed on a verification timeout was fixed.
- getaddressutxos is now limited to the requested height range and limit (an external contribution).
- Some long state tests were converted to regtest to cut CI time.
- The snapshot page (snapshots.zfnd.org) and the book documentation were published, Mergify was replaced with GitHub's native merge queue, and the more than 120 labels were cleaned up to about 20.
- Next up are a fix for the sync restart loop near the tip, mining and mempool performance PRs, an external contributor's PR for shielded bundle verification caching, and Zebra's side of the new P2P v2, which is at the draft ZIP stage.
ZODL: core libraries and Zallet update
ZODL · Pacu, Kris Nuttycombe, Alfredo
- The team worked a lot on coinholder polling in collaboration with Valar Group.
- It released Zallet beta 3, which fixed a migration path problem reported by miners and some items from the Least Authority audit.
- The audit has two rounds, and it is now in the second round, which confirms fixes and takes suggestions.
- GitHub security advisories were flooded with plain bug reports and were cleaned up, and the team asked people to check whether something is a security issue before submitting.
- Beta 4 aims to ship more often with fewer changes.
- Some miners are already using Zallet all the time.
- The team asked people to back up zcashd wallet.dat files that manage keys in various ways, import them and file issues, and it is cleaning up the migration path documentation to prevent confusion between wallet.dat and wallet.db.
- Because the mobile SDK versions are a mess, a cleanup release is coming (Kris corrected that 3.0 is still to come). Coinholder polling will be included in the upcoming SDK.
- The team is also reviewing the librustzcash PRs for Tachyon and v7 and the P2P v2 ZIP.
- When Alfredo asked about the performance improvements in the core libraries that Valar Group forked, Kris answered that there is an ongoing process by which Danny Williams is backporting them to the Zcash organization's repositories.
- The team asked for third-party review of four pasta_curves crate PRs (103, 106, 107 and others).
Zaino update
Zingo Labs · Eli, Pacu
- Eli, who joined two months ago, said the work of modularizing the internals of index construction is nearly done.
- In the new architecture each index is optional, so by default only the compact block index is built, and computationally expensive indexes such as gettxoutsetinfo are turned on only when needed.
- Arlo merged and released all of the mempool and security-related fixes.
- The team built a new test harness and observability tooling that runs a full sync test every release cycle and measures blocks per second and CPU flame graphs, and plans to make some optimizations over the next month or two.
- Pacu asked whether compact block data, which Zallet does not need, can be turned off, and Eli answered that each index has a configuration section, so compact block indexing can be turned off too.
- On the possibility of providing snapshots, Eli said that entangled indexes must be snapshotted together, but compact blocks can be snapshotted alone.
- Pacu said once a week would be enough rather than daily.
Network Sustainability Mechanism (NSM) update
Shielded Labs · Judah
- The team is waiting for the NU 7 polling results.
- The alternative ZIP 234 draft that keeps halvings needs additional review.
Crosslink and dynamic fees update
Shielded Labs · Jason
- Last week the team met in person for project planning and the roadmap, and held a workshop.
- It also announced the plan for the next ZEC payout of the incentivized testnet.
- Nate is absent this week and will give a detailed update on the next call, in two weeks.
- Everyone on dynamic fees was also away, so they agreed to give a more concrete update on the next call.
ZODL formal verification update
ZODL · Kris Nuttycombe
- Formal verification work continues, and Daira Emma plans to publish an update blog post soon.
- The balance and validity games were formalized and extended to several items beyond the circuit, but the speaker said they did not know the exact details.
Open presentation: ZIP 316 revision 2 and a USD-based minimum fee proposal
ZODL · Kris Nuttycombe, Pili, Zooko
- Kris shared the implementation PR for ZIP 316 revision 2.
- It is important for expressing standalone account-level viewing keys, which are needed for importing zcashd wallets, and it is nearly ready to merge.
- Kris also asked for feedback on the draft PR for ZIP 248 (the extensible transaction format).
- Since everything is going into bundles anyway, Kris is considering creating a coinbase-only bundle type to clean up the old approach of recording block height and miner data in a fake input of the transparent bundle, and also adding miner signaling.
- ZIP editors discussed Shielded Labs' PR to lower the ZIP 317 marginal fee by 1000 zat.
- Kris proposed proceeding with it, but having full nodes use an algorithm that sets a USD-based minimum fee from a lagged average price history, so that ZIP editors are excluded from fee changes.
- Kris believes that full blocks cannot be tolerated right now and that Shielded Labs' dynamic fees do not define a stable control algorithm.
- Pili said they had doubts about pegging to USD.
- Zooko suggested not debating this on a public call but running a meta-process in which different wallets implement different approaches and experiment, and also proposed a design review by an independent mechanism economist.
This summary is based on the original meeting notes; check them for exact details.