HackMD
Final Dev Update
Contributions to core development in Ethereum
Project Abstract
I worked on improving the Nimbus consensus client. Most of my workload throughout this fellowship has been dedicated to transitioning the block processing method to a model where blocks are fully indexed in memory, with the storage adopting an append-only file implementation.
Status Report
The initial plan centered around improving bip32, as proposed by Paul Miller. However, plans shifted, leading me to Nimbus-eth2 (An Ethereum Consensus Client in Nim). I had to learn the Nim programming language and understand the Ethereum Consensus. This project consisted of three major components:
-
Append-only file implementation: Initially challenging due to my unfamiliarity with distributed systems. After considering an SQLite write-ahead log model, implementation progressed steadily, addressing issues like understanding vfs, parameter types, necessary macros, and overcoming linking issues.
-
Fully indexing block in memory: Relatively straightforward, providing insights into indexing and tree data structures. However, working with the
openArray[byte]datatype posed challenges, only functioning when used as a proc parameter. Despite this, the task contributed significantly to my understanding of the codebase. -
Removing client dependencies on SQLite: Now, I am presently focused on using what I have directly to work in the Nimbus consensus client.
Future of the Project
There isn't much else to be done with regards to this log-based structure for hot data, but the goal is to make sure the eventual goal is achieved. However, there is more in the grand scheme of things. I had an initial plan at the beginning of my proposal to working on the consensus client. Since I have recently been dabbling into understanding tries, I thinking helping out my other fellows on the Verkle trie would help me keep working on the distributed systems alongside Cryptography. But also, here are some of the other contributions I intend on working on for the next few months.
- Implement instant one-shot syncing based on the zero-knowledge proofs from the DendrETH project.
- Implement a distributed validator setup in the Nimbus beacon node, reducing the risk of validator downtime.
- Implement a stateless mode for the Nimbus execution client
- Create a single Nimbus binary combining our consensus and execution clients,
and many more
Self-Evaluation
It has been an wholesome experience. I do wish I had known some things or had done some things sooner. I had to deal with the conflict of asking questions when it came down to it, but in the end, you'll be shocked about people's willingness to help. It makes it seem more like a project towards a goal than a show of skills. The discord and stand up calls were fantastic. different characters. different minds. different approach. Overall, I liked the project I was working on, and I like where I am headed. I look forward to what I could do
Feedback about EPF
EPF's organizational aspects, including the AMA sessions with ecosystem veteran developers, were commendable. Grateful for the opportunity, the fellowship has been a positive experience, shaping my direction and skills. The fellowship will continue to grow, and perhaps, by cohort five, I would be assisting the future fellows as I have seen from a couple alumnis of the program
