Chain interface: drying up and splitting it in multiple ones

[ariard]

Once #16426 merged, we should keep drying up the interface to make it fully asynchronously, few ideas to brainstorm on:

• all getHeight, getBlockHeight and findFirstBlockWithTimeAndHeight-style methods should be removed once rescan logic moved in its own thread behind the interface (I have already a functional branch for this)
• all mempool related methods (mempoolMinFee) could be moved in its own Mempool::interface, client could care only about fee estimation, make it asynchronous too
• initMessage, initError and others could be moved in Node::interface
• handleRpc and others, wallet should have its own RPC server

I think last two issues could be started today, they don't rely on any other PRs.

We may also have a Broadcast interface, tracking the chain state and broadcasting a tx are really 2 different logics.

@ryanofsky pretty sure you have thoughts on it too.

[ryanofsky]

Things suggested above sound good to me. There's a TODO comment with similar suggestions.

• initMessage, initError and others could be moved in Node::interface

The interfaces::Node class is intended more for the gui and would give the wallet more functionality than it needs but I've previously thought these could go in a small interfaces::Ui or interfaces::WalletClient class along with the loadWallet notification (#15288 (comment))

[maflcko]

Is this still an issue?

[ariard]

@MarcoFalke

From a quick read, I would say the suggestion about initMessage and initError is still relevant against today's code (a). The other one about isolating the rescan logic on the wallet-side is still relevant (b). About a "Mempool::interface" I think some information could be relevant in the recent discussions about mempool-code reorganizations (c). Wallet should have its own RPC server is already documented as a TODO in doc/multiprocess.md iirc (d). A new Broadcast or extended Chain interface would be relevant for AltNet (e).

I don't have interest in doing a) or b) or d) in the coming future. I do have strong interest in c) and e) soon.

From an issue-management standpoint, feel free to close the issue or re-dispatch the relevant informations elsewhere.

[ryanofsky]

I think a lot of what's discussed here is still relevant, but there's not really a need to keep the issue open. I think any new ideas that we don't want to forget about could be added to the TODO comment. Otherwise I don't think there's any decisions that need to be discussed or ongoing development that needs to be tracked. The chain interface is already a lot simpler than it used to be after #16426, and it could just be broken up a little more.