Madara Community Call #3

[AbdelStark]

ZOOM LINK

Agenda

• Backlog review and update on tasks by contributors
• Sharingan testnet update (@abdelhamidbakhta @tarrencev)
• RPC implementation deep dive (@tdelabro)

Information

• When: Fri, 7 Apr 2023 at 2p.m CEST

Paris, France Fri, 7 Apr 2023 at 14:00 CEST
New York, USA Fri, 7 Apr 2023 at 08:00 EDT
UTC, Time Zone Fri, 7 Apr 2023 at 12:00
Sydney, Australia Fri, 7 Apr 2023 at 22:00 AEST

• Where: on Zoom

"Wherever There Is Light — There Are Always Shadows."

[AbdelStark]

RPC implementation deep dive

1. Introduction

During the third community call, the organizers presented a deep dive into the implementation of the Full RPC (Remote Procedure Call) Client for Starknet sequencer. This document provides a summary of the key points discussed during the call, including the motivations behind the RPC Client, an overview of its implementation, and potential challenges and future improvements.

2. Motivations for the RPC Client

The RPC Client aims to provide full RPC support for Starknet, allowing users to interact with the blockchain through various RPC methods. The goal is to achieve full coverage of the Open RPC specification for Starknet while ensuring compatibility with Substrate.

3. Overview of the RPC Client Implementation

3.1. Retrieving Block Hashes

The RPC Client retrieves block hashes by querying the current block and using the Substrate Block Hash to index into the storage at a specific block. This method is preferred over directly accessing the storage because it provides a more flexible and less risky approach. It also does not impact the performance of the actual blockchain.

3.2. Runtime API and Storage Overrides

To optimize performance, the RPC Client first attempts to read from the storage directly using storage overrides. If this fails, it falls back to querying the runtime API. This approach reduces the overhead of switching between client execution and the WebAssembly runtime when reading from the storage.

3.3. Substrate and Starknet Block Hash Mapping

Since Starknet block hashes and Substrate block hashes are different, the RPC Client uses a reverse mapping from Starknet block hashes to Substrate block hashes. The client listens for new finalized blocks and maintains a mapping database, which allows it to quickly retrieve the Substrate block hash associated with a given Starknet block hash.

4. Challenges and Future Improvements

Implementing new RPC methods will be an interesting task for new contributors. Some challenges may arise along the way, and innovative ways to read from the storage will need to be found. Additionally, implementing a caching mechanism, such as LRU, can help optimize performance if many users call the same RPC methods.

5. Conclusion

The RPC Client is a crucial component for providing full RPC support for Starknet. The implementation is set up to make it easy for new contributors to get involved and add more RPC methods. Timothy is available to help anyone interested in contributing to this project. The community is encouraged to work together to achieve full coverage of the Open RPC specification for Starknet.

[AbdelStark]

Summary of the Third Madara Community Call

• Chapter 1: Introduction
  Timestamp: 0:00-0:15
• Chapter 2: Backlog Review
  Timestamp: 0:16-2:07
  Contributors give updates on their tasks, including integration tests, migration scripts, and optimizing block processing speed.
• Chapter 3: Demo of Madara Benchmark
  Timestamp: 2:08-5:08
  Discussion of benchmarking and improving block processing speed.
  Talk about storing state tree and first noun, and optimizing computation.
• Chapter 4: Deep Dive of RPC Implementation
  Timestamp: 5:09-10:22

Discussion of emitting events and defining traits for querying data.
Discussion of FPC method standard and automatic generation of code.
Mention of Ink smart contract language and ability to emit events from an extrinsic.
Chapter 5: Substrate Architecture Overview
Timestamp: 10:23-16:09
Discussion of updating storage and depositing events.
Talk about defining traits for querying data and using a mapping sync trait to implement a stream.

• Chapter 6: Q&A and Discussion
  Timestamp: 16:10-End
  Encouragement for new contributors to add to the discussion or ask questions.
  Mention of public test net and running dollar game, optimization track, and gas accounting.
  Follow-up tasks mentioned: implementing a new RPC, updating the Debit database, adding more accounts, and rewriting different don't.

Action items:

• Backlog review: Each contributor gave an update on their tasks.
• Public test net and running Dojo game: Tarrence and Loaf explained the goal of running a public test net and running dollar game on top of it.
• Optimization track: Discussion of the blockifier and moving the global state of computation to the client side.
• Gas accounting: Brief discussion of how it prices the cost of an operation.