Node API: make chunk encoding explicit #1278
Conversation
| // [KZG proof: 32 bytes] | ||
| // [Coeff 1: 32 bytes] | ||
| // [Coeff 2: 32 bytes] | ||
| // ... | ||
| // [Coeff n: 32 bytes] | ||
| // | ||
| // The KZG proof is a point on G1 and is serialized with bn254.G1Affine.Bytes(). | ||
| // The coefficients are field elements in bn254 and serialized with fr.Element.Marshal(). |
There was a problem hiding this comment.
This is missing a few parts. How about this?
// [version] 1 byte currently, only version 1 is supported
// [chunkLength] 7 bytes little endian
// Chunk 1:
// [proof] 32 bytes
// [coeff 1] 32 bytes
// [coeff 2] 32 bytes
// ...
// [coeff chunkLength] 32 bytes +-------------------------------------------------+
// Chunk 2: | point on G1 |
// [proof] 32 bytes <----| serialized via bn254.G1Affine.Bytes() |
// [coeff 1] 32 bytes | github.com/consensys/gnark-crypto/ecc/bn254 |
// [coeff 2] 32 bytes +-------------------------------------------------+
// ...
// [coeff chunkLength] 32 bytes
// ...
// Chunk n: +-------------------------------------------------+
// [proof] 32 bytes | field element in bn256 |
// [coeff 1] 32 bytes <--- | serialized via fr.Element.Marshal() |
// [coeff 2] 32 bytes | github.com/consensys/gnark-crypto/ecc/bn254/fr |
// ... +-------------------------------------------------+
// [coeff chunkLength] 32 bytes
//
// The size of each individual chunk (proof+coeffs) is 32*(1 + chunkLength) bytes. If the total length of the GNARK
// encoded bytes is S, the number of chunks (n) is equal to (S-8) / (32*(1 + chunkLength)) bytes.
There was a problem hiding this comment.
Ideally, this documentation would not use golang code as part of its specification (i.e. we should look at things like fr.Element.Marshal() and write town the byte format in the doc). But this is at least better than not having any documentation at all. Later when we have more time, perhaps worth expanding the docs in this way.
There was a problem hiding this comment.
I don't think version etc are in chunk here, they are only in bundle / chunks in db, not when it's returned
This is the encoding for one chunk, there is no "chunk 1, 2...n" here though
The golang is the only one we have, if there are others, we can put them in the list here
Why are these changes needed?
Make it clear / explicit about the chunk encoding format for clients of the Node API.
Checks