← All calls
ARBORIST#35

ZIP 317 phased rollout roadmap

A roadmap for rolling out the ZIP 317 fee overhaul in stages, in the order wallets, relay, consensus, was discussed.

Agenda summary 9

  1. Zebra update - beta.14 and a wallet feature question

    Zcash Foundation · Marek, Deirdre, Dodger

    • Zebra beta.14 was released a week ago. CI works stably, and the start command became the default when running zebrad.
    • Sync became faster after the team fixed a problem of updating the tree state twice per block once the checkpoint range ended. A lightwalletd sync regression caused by deserializing transactions on rayon threads was reverted.
    • The team is working on making the release process for Zebra release candidates smooth. A new engineer, Arya, joined.
    • In response to Adi's question, Dodger answered that building a wallet into Zebra itself is not on the roadmap. Currently the focus is on alternative lightwalletd support and getblocktemplate.
    • There are also no plans to build in a miner; the plan is to let miners get block templates from Zebra via getblocktemplate and mine with their own software.
  2. Zcashd update - sync improvements and memory usage

    Electric Coin Company · Steven, Daira

    • The node side improved considerably with 5.1.0 and 5.2.0, and the focus is now on improving wallet sync speed.
    • There were reports that memory usage increases during sync and some users hit OOM. The team succeeded in reproducing it and found an initial fix to reduce peak memory; the cause appears to be increased throughput rather than a leak.
    • Daira said OOM also occurs for users who do not scan shielded addresses, so there may be other factors. Cases were also confirmed where steady-state usage exceeds 9GB on an 8GB VM.
  3. Fee mechanism change - ZIP 317 discussion

    Electric Coin Company · Steven, Nate, Daira, Adi, Str4d

    • On ZIP 317 (PR 631), posted by Adi of Nighthawk, Nate said ECC is broadly supportive, on the premise that parameters will be adjusted. It will incorporate user research and on-chain analysis, and they see it as part of long-term fee improvements.
    • Greg, a new ECC engineer, is analyzing how much past transactions paid and how much they would pay under the fee change. The aim is to check the impact on ordinary use with data.
    • Nate proposed splitting the ZIP 317 rollout into stages. Wallets implement the rules first, then relay rules change, and neither needs a consensus change, so it can be done quickly. They said extensions such as congestion-linked fees and the roadmap should be handled separately.
    • Daira said transparent wallets already adjust fees to chain conditions, so there is no big problem. However, they acknowledged that some wallets using the existing fixed fee of 10,000 zat have a problem when it exceeds 10 times the default fee.
    • Nate said fees proportional to inputs and outputs affect various use cases such as mining pools and exchanges, so we should not assume only a typical user. Adi saw shielded miner transactions as having around 50 to 100 inputs and outputs, so not a big change.
    • Daira expected low-fee transactions will eventually be filtered out by relay limits. Nate pointed out UX issues such as whether a wallet can learn of the rejection.
  4. ZIP 317 - deployment section and transaction expiry height

    Electric Coin Company · Str4d, Dodger, Kris

    • Str4d relayed that the ZIP editors made several comments during the ZIP 317 review asking for the deployment section to be filled in, and these have not yet been addressed.
    • The default expiry height of the transaction builder ECC maintains is 40 blocks. Some Bitcoin-derived builders fix this field at 0, and some implementations cannot parse transactions with an expiry height set.
    • Kris said Greg's analysis is also collecting statistics on the expiry heights actually set in real transactions.
  5. ZSA fee proposal

    QEDIT · Daniel Benarroch

    • A fee proposal for ZSA is being written and will be posted within one to two weeks. Sapling and transparent are left as they are, and only ZSA is priced differently.
    • The proposal is to lower the price per byte and per action to encourage ZSA use, and to charge issuance a fixed fee per issuance and a small fee per output.
    • The aim is to keep ZEC as the core currency and prevent junk assets. The team floated raising the issuance fee to around 1 ZEC and sending it to a posterity fund, the Zcash Foundation, the Grants Committee fund, or similar.
    • Adi said ZSA will come to mainnet next year or later, so the priority is still ZIP 317, and encouraging ZSA and preventing spam should be considered together. They also said fee changes need to be coordinated and rolled out across wallets, exchanges and even hardware wallets.
  6. Mobile SDK sync improvement roadmap

    Electric Coin Company · Steven, Str4d, Daira, Adi

    • ECC's top engineering priority is fast sync for the mobile SDK, with Edge, Unstoppable and Nighthawk as the main beneficiaries. Improvements will be rolled out in several stages.
    • It depends on internal testing, but the goal is to release the iOS SDK first and Android shortly after, aiming to package the first iOS update next week.
    • The next SDK includes the unified address (UA) migration and parallelized trial decryption. A change to the commitment tree and witness update method follows, and full DAGSync comes after that because the SDK interface may change.
    • The parallel trial decryption code exists and was integrated into Sapling, but Str4d did not remember whether it was applied to Orchard as well. The SDK does not support Orchard yet, so there is currently no impact.
  7. UA and Orchard rollout discussion

    Electric Coin Company · Str4d, Daira

    • In the UA migration, the spending key backend includes Orchard keys, but when spending the SDK picks Sapling rather than Orchard. Trial decryption of Orchard outputs will be added in a later update.
    • Daira pointed out that a UA containing an Orchard component must not be issued before Orchard can be received, and the team will file issue 630 in librustzcash to verify this.
    • SDK wallets ignore the Orchard component of a UA and send to the Sapling component. Orchard-only UAs created by Zecwallet Lite will not work in the next SDK.
  8. ZSA progress - circuit size and value representation

    QEDIT · Daniel Benarroch, Daira, Str4d, Nate

    • The team will restructure and submit the ZSA proposal's milestones divided into design and implementation, and deployment and long-term testing, and said it will consult Adi of the Grants Committee on the procedure.
    • Circuit implementation is about halfway done, and the circuit size may exceed the target limit, so the constraints may need to be reconsidered. Daira asked how much it exceeds 2^11 but there were no concrete figures yet.
    • The costly part is the added Sinsemilla hash rounds for attaching the type. Daira pointed out that fixed-base multiplication becomes variable-base, and that the variable-base multiplication implementation has a lot of room for improvement.
    • Str4d said value representations beyond u64 may be needed because of Ethereum bridges (wei, ERC20's u256). Daira answered that u192 is also possible and the bridge can require the low digits to be zero.
    • Daira pointed out that the explanation of dummy handling is missing from the ZIP circuit section. For a split, V old must be enforced to be non-zero to keep the Merkle path verification, and Str4d also said it should be made clear in the ZIP.
  9. Unified Spending Key and Unified Viewing Key

    Electric Coin Company · Hazel Ohearn, Kris, Str4d, Dodger, Daira

    • Hazel asked why the USK and UFVK are not protocol-level concepts and are doc(hidden). Kris answered that the UFVK has its encoding defined in ZIP 316, while the USK has only the derivation method in ZIP 316.
    • The USK encoding PR for crossing the FFI boundary will be merged soon, and it is just an internal convenience for wallets, not for backup or transfer. It is for environments that use only ASCII strings, like React Native, and is protected behind an unstable feature flag.
    • A UA is given to counterparties who send funds and a UFVK to the few who will be given viewing access, but the USK, the root of spending authority, is not exchanged. Recovery is handled by the mnemonic seed. Kris compared it to not copying an SSH private key to another device.
    • Dodger and Daira said there may be situations where hardware wallets support only some pools. Str4d said that when a new pool appears every USK would have to be updated, so passing around USKs is the wrong feature.

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