Currently the C1 phase API call requires a path to the replica file. Challenges are generated and the corresponding Merkle paths are extracted from the replica together with its Merkle tree.
The idea is to create a new API call, that doesn't require the full replica file, but only the pieces that are needed for the Merkle paths. This means that the challenges can be generated externally (the process is deterministic) and the corresponding pieces of the replica file are extracted and then transmitted to the new API call.
The execution would be the same as for the normal C1 phase, the only difference is that the lowest level of the replica Merkle tree (the replica itself) isn't used, but a list of the data that resides at the corresponding challenges.
Acceptance criteria
There is a new API call that works.
Risks + pitfalls
This might need changes to the merkletree crate as it's abstraction of the Merkle Trees we use.
Where to begin
Check if the replica is part of the Merkle path generation directly in merkletree or if the replica offsets are already extracted separately
Depending on that, either implement it directly in rust-fil-proofs or modify merkletree to support such "sparse leafs"
Description
Currently the C1 phase API call requires a path to the replica file. Challenges are generated and the corresponding Merkle paths are extracted from the replica together with its Merkle tree.
The idea is to create a new API call, that doesn't require the full replica file, but only the pieces that are needed for the Merkle paths. This means that the challenges can be generated externally (the process is deterministic) and the corresponding pieces of the replica file are extracted and then transmitted to the new API call.
The execution would be the same as for the normal C1 phase, the only difference is that the lowest level of the replica Merkle tree (the replica itself) isn't used, but a list of the data that resides at the corresponding challenges.
Acceptance criteria
There is a new API call that works.
Risks + pitfalls
This might need changes to the
merkletreecrate as it's abstraction of the Merkle Trees we use.Where to begin
merkletreeor if the replica offsets are already extracted separatelyrust-fil-proofsor modifymerkletreeto support such "sparse leafs"