Few questions about adding support for the BSC network on a forked branch of Reth

[unclezoro]

hi, Reth team,I am a developer from the BNB Chain team. Our interest in the Reth project was piqued when we started working on the Akula project. The recent performance report and roadmap blog of Reth is quite impressive to us. As a result, we attempt to add support for the BSC network on a forked branch of Reth. We've encountered some high-level issues during our design process, and we'd like to consult with your team to ensure our implementation aligns with the best practices for modularity.

Here are the questions:

1. At present, the 'stage sync' or the 'insert block' of blockchain trees are triggered by external programs such as the op-node or the consensus client. However, when we look at BSC or other EVM blockchains, the scenario is slightly different as they dont have these “sidecar” programs. Their block syncing is triggered by modules like the 'downloader' or the 'block fetcher' that sit atop the P2P network module. We could introduce a similar program that feeds the blockchain data, or we implement a different “downloader”. For the latter case, we didn't find any patterns to follow, nor did we know how to implement modular design in accordance with 'reth'，appreciate if any advice.
2. The implementation of Optimism currently does not support stage sync. Will there be future updates that enable stage sync for Optimism? When considering sync modes for other networks, which one we should refer to? Ethereum or Optimism?
3. The definition of consensus traits is rather simple. The “finalize” logic, which originally belonged to the consensus, has been shifted to the executor, as seen here. However, the “finalize” of BSC is quite heavy, check here, we have to put it inside the “EVMProcessor”, make it quite fat. From the perspective of module cohesion, we hope that these logic can be incorporated into the consensus engine, but the consensus trait is so simple that we can hardly achieve this, any better idea?

[joshieDo]

re 2) As of yesterday it should be supported #7552

[unclezoro]

Really good to know. Thanks @joshieDo

[mattsse]

hey @unclezoro , very excited about this and happy to help, feel free to ping me on tg @mattsse as well or a new channel.

for p2p we have this block import stub:

reth/crates/net/network/src/manager.rs

Lines 85 to 86 in d852f7f

	/// Handles block imports according to the `eth` protocol.
	block_import: Box<dyn BlockImport>,

which currently doesn't do anything in PoS, but this is how we envisioned block import for p2p would look like, happy to change things if that makes it more convenient for you.

bypassing the engine API would also be possible I assume, we made some progress on abstracting the launch process which might come in handy, but I'll probably need more context about your exact requirements, is there code I can look at?

[unclezoro]

thanks for the quick response. Good to know the block_import stub, will check this.

Regarding the consensus engine, here is the POC code:

Finalize in EVMProcessor: https://github.com/bnb-chain/reth/blob/ed1475376fa138b315f86639df7135f0d59cad22/crates/revm/src/processor.rs#L682
Parlia consensus: https://github.com/bnb-chain/reth/tree/ed1475376fa138b315f86639df7135f0d59cad22/crates/consensus/parlia/src
BSC processor: https://github.com/bnb-chain/reth/blob/ed1475376fa138b315f86639df7135f0d59cad22/crates/revm/src/bsc/processor.rs

[unclezoro]

hi @mattsse I can not ping you privately through TG because of your security setting I guess, this is my TG https://t.me/zorowhite, appreciate if you can add me as contact.
I think #2 and #3 is quite clear to us. Regarding #1, the present auto seal consensus is limited to a solitary node. Would the team consider to extend it to a private network? Like a single auto seal miner is active, with other full nodes able to synchronize through p2p network. This could serve as a good example for us to broaden support to other networks like BSC, which does not have any consensus client or op-node.

PS: EIP225 Clique consensus protocol, if reth could implement such consensus, it will make things even simple.