← All calls
ARBORIST#34

Responding to network load and ZIP 317

ECC shared the status of its response to the surge in network load, and the draft ZIP 317 fee proposal was introduced.

Agenda summary 6

  1. Zebra update - preparing a release candidate

    Zcash Foundation · Teor

    • A few weeks ago the Zebra team lead was replaced by Teor. Everyone on the team helped run the conference, so development was slow this month.
    • For the release candidate, the team did test cleanup, dependency updates, minor performance improvements and small fixes to the network protocol.
    • The previous release had breaking configuration changes and several performance improvements, and the goal is to prepare a release candidate within the next few releases.
  2. Zcashd update - impact of the network load and the response

    Electric Coin Company · Steven

    • Shielded transaction load rose sharply from mid-June. Node operation and P2P were little affected apart from slower block propagation, while miners and exchanges using full node wallets saw longer scan times and more orphan blocks.
    • Mobile wallets were hit hardest. The SDK-based Edge, Unstoppable and Nighthawk had very slow or failed syncs, while Ywallet and Zecwallet Lite generally worked well.
    • The load shifted from Orchard-centered to Sapling-centered in mid-June. Transactions that averaged just over 100 outputs now have over 1000 outputs, so the number of transactions fell but the verification burden is similar.
    • The chain size roughly tripled after mid-June, from about 30GB to about 83GB. 5.1.0 cut Sapling and Orchard block verification time by 80%, and 5.2.0 included shorter propagation times and parallelized full node wallet scanning.
    • Per partner feedback, the orphan block rate came down from 8% at one point to 1.8 to 2.8%. Large zcashd wallets with hundreds of thousands of addresses scan slowly but operation is not affected.
  3. Zcashd update - mobile SDK sync and ECC's direction in response

    Electric Coin Company · Steven

    • Among the mobile SDK's FFI layer, sync improvements and API enhancements, fast sync was raised to the top priority. It is still before development and benchmarks, but the expectation is to cut sync that took hours to days to under an hour.
    • ECC makes engineering its top priority while also informing ecosystem partners of the situation and identifying remaining node-level problems.
    • Research staff are moving from Proof of Stake to fee mechanisms, and Nate will review EIP 1559-style ideas and also look at Adi's ZIP proposal.
    • It said any change, fees included, will be evaluated comprehensively for side effects on mobile wallet UX and the mempool. The growth team will run a user survey and share the results, and in a second stage plans to find and resolve long-term vulnerabilities in advance.
  4. Open discussion - wallet SDK integration and the unified address transition

    Electric Coin Company · Daira, Str4d, Steven, Dodger

    • Work to unify the librustzcash fork the wallet SDKs used onto the latest version is nearly done. Performance work proceeds on top of it, and the long-stalled zcash_client_backend crate release has also come closer.
    • The new mobile SDK is the release in which wallets switch to unified addresses. Viewing keys migrate to unified viewing keys containing transparent, Sapling and Orchard.
    • The first SDK release does not support Orchard spending, so Orchard can be dropped from the UA's address components for now. The spending key includes Orchard so the migration needs to happen only once, and Orchard support follows right after.
    • Steven mentioned the lesson that it would have been better to release unified addresses first and Orchard in a later upgrade. The ECC wallet also plans to use the lightwalletd architecture as is.
  5. Open discussion - detection keys and DAGSync

    Str4d, Kris, Daira, Dodger, Pacu

    • Str4d explained that the bottleneck of steady-state sync is detecting new spends on the chain that are not linked to the existing wallet, that is, trial decryption of every output. Big improvements come from the detection key approach or from receiving transactions directly without scanning the chain.
    • Detection keys require sender compatibility and an address change, so they are hard to deploy alone. They can be layered on top once unified addresses are established.
    • In Str4d's DAG sync idea, newly discovered funds are incorporated into the wallet's graph, so even on seed recovery only recent transactions need to be found, without a full linear scan. Daira pointed out that outputs remain on-chain for seed recovery.
    • Dodger put forward an idea in which a full node scans on behalf of users with detection keys and light wallets connected over Tor split up ranges to delegate. An opinion was raised to hold a workshop, which would also reduce light wallet server centralization.
    • Pacu said wallets do not actively watch the mempool, so there is a lot of room to improve the perceived speed of transactions. Str4d answered that shielded change cannot be spent before it is mined, so the mempool matters mainly for transparent transactions.
  6. Open discussion - pausing shielded transactions, block size and ZIP 317

    Steven, Str4d, Daira, Dodger

    • On the question of temporarily pausing shielded transactions, Str4d said blocks can be filled with transparent transactions too, so it would have no effect and would only reduce proof verification costs. Daira said it would not work because it achieves the attacker's goal.
    • Daira presented a proposal, as a miner option rather than a consensus rule, to halve the effective block size, which roughly halves chain growth and verification time. It is already possible as a zcashd option.
    • Steven worried that shrinking blocks could crowd out ordinary shielded payments. Daira acknowledged that an estimate made when reducing block time at Blossom may have been wrong, since even after parallel verification was implemented it was close to the limit.
    • Dodger saw it as reasonable as a temporary measure but, since the goal is much larger transaction volume in 5 to 10 years, saw this episode as a trigger forcing improvements. Steven named transactions or block aggregation based on Halo 2 recursion as the long-term solution.
    • A draft ZIP 317 (PR 631) from Nighthawk's Aditya, proposing a fee proportional to the number of inputs and outputs, has been posted, and Daira asked people to leave comments. Nate will review it together with EIP 1559-style research.

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