Skip to content

Implement fee source#29

Merged
febyeji merged 11 commits into
febyeji:cbf-chain-source-cleanupfrom
randomlogin:cbf-wiring-up
Jun 12, 2026
Merged

Implement fee source#29
febyeji merged 11 commits into
febyeji:cbf-chain-source-cleanupfrom
randomlogin:cbf-wiring-up

Conversation

@randomlogin

Copy link
Copy Markdown
Collaborator

We implement three fee sources for the CBF chain source.

  1. esplora
  2. electrum.
    For electrum we altered the existing code from the electrum chain
    source to make it reusable for the CBF node We change runtime argument
    to start() to be a field in CbfChainSource struct, because
    electrum fee source explictely needs it.
  3. CBF native fee source.
    It downloads blocks and calculates fees based on previous blocks. We
    maintain a cache up to BLOCK_FEE_CACHE_CAPACITY=14 blocks. For
    confirmation target we use estimation vie percentil of previous
    blocks' fees. Also if we download a block on a matched block filter,
    we insert it in the cache to avoid re-downloading.
    As the block download might be slow, we don't fail the fee estimation,
    but rather use fallback fees. This is especially relevant to the very
    first call of update_fee_estimates during the start.

We add helper pure functions to a new file src/util.rs.

randomlogin and others added 9 commits June 3, 2026 01:06
Add stub methods/functions, add basic build and start of the CBF chain
source as well as basic struct containing the fields which undoubtedtly
are needed.
Previously tests assumed that the chain source of the lightning node and
is node which mines. This is not the case with CBF chain source which
needs to wait until after mining a new block a new tips propagates to
it.

`wait_for_block` is made to return a new height and a new function
`wait_for_node_tip` is added which waits until the given height is
processed (returned via `status.best_block` ) on a given node.
Ask wallet for revealed spks, register them. Implement `Listen` trait
ans add register_script method as well as implementation of registered
scripts/outputs.
When `process_kyoto_events` processes events it decides whether we need
to take any action (e.g. apply block). These actions are sent to a new
abstraction — `BlockApplicator` which holds `ChainListener`
(which in turn has wallets and can apply blocks / filtered blocks). This
`BlockApplicator` has to have a receiver of a channel (and
`process_kyoto_events` has to have a sender to this channel. Thus we
cannot create them in `new`, because we would own them at
`CbfChainSource`, so they are created in `start`.

Also this commit ran `cargo fmt --all` which was missed previously.
Co-authored-by: febyeji <yeji.han@sf.snu.ac.kr>
Added 4 variants of `ChainOp`:
- ConnectFull,
- ConnectFiltered,
- Disconnect,
- Synced

Now `process_kyoto_events` reacts to an event from kyoto and sends a
`ChainOp` to listener (`BlockApplicator`).

Note that on a filter with no match we still need to apply header and we
need double check that header in the canonical chain is the relevant
one and has not been reorged. Right now it is left as a todo, because of
an upstream PR.

Co-authored-by: febyeji <yeji.han@sf.snu.ac.kr>
We implement three fee sources for the CBF chain source.
1) esplora
2) electrum.
  For electrum we altered the existing code from the electrum chain
  source to make it reusable for the CBF node We change runtime argument
  to `start()` to be a field in `CbfChainSource` struct, because
  electrum fee source explictely needs it.
3) CBF native fee source.
  It downloads blocks and calculates fees based on previous blocks. We
  maintain a cache up to `BLOCK_FEE_CACHE_CAPACITY=14` blocks. For
  confirmation target we use estimation vie percentil of previous
  blocks' fees. Also if we download a block on a matched block filter,
  we insert it in the cache to avoid re-downloading.
  As the block download might be slow, we don't fail the fee estimation,
  but rather use fallback fees. This is especially relevant to the very
  first call of `update_fee_rate_estimates` during the start.

We add helper pure functions in a new file `src/util.rs`.

Co-authored-by: febyeji <yeji.han@sf.snu.ac.kr>
@randomlogin randomlogin requested a review from febyeji June 9, 2026 13:32
Co-authored-by: febyeji <yeji.han@sf.snu.ac.kr>
@randomlogin randomlogin mentioned this pull request Jun 9, 2026

@febyeji febyeji left a comment

Copy link
Copy Markdown
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This seeds the CBF watch set only with scripts that were already revealed at start. Can't new internal/change scripts be revealed later when the wallet builds transactions?

Onchain wallet's scripts are pulled each time we receive `IndexedFilter`
event. That way we rely on a single source of truth (wallet) instead of
having two overlapping script sets.

Now we pass a reference to the `onchain_wallet` to make possible get all
revealed scripts from it.

Co-authored-by: febyeji <yeji.han@sf.snu.ac.kr>
@randomlogin

Copy link
Copy Markdown
Collaborator Author

Changed the logics to have just LDK's scripts in registered_scripts. BDK's scripts are retrieved on-the-fly each time we need them (when we receive IndexedFilter event).

@randomlogin randomlogin requested a review from febyeji June 11, 2026 20:49

@febyeji febyeji left a comment

Copy link
Copy Markdown
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM.

@febyeji febyeji merged commit d2b769f into febyeji:cbf-chain-source-cleanup Jun 12, 2026
12 of 23 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants