update Deneb for blob sidecar inclusion proofs #5565
Conversation
`BlobSidecar` is no longer signed, instead use Merkle proof to link blobs with block. - ethereum/consensus-specs#3531 Associated beacon-API / builder-specs still TBD; minimal changes done to compile in similar style to previous spec, but not standardized yet. - ethereum/beacon-APIs#369 - ethereum/builder-specs#90
etan-status
force-pushed
the
dev/etan/df-proof
branch
from
411bfde to
9cc745a
Compare
etan-status
force-pushed
the
dev/etan/df-proof
branch
from
998d1db to
a48ca2b
Compare
| # [REJECT] The sidecar's blob is valid as verified by `verify_blob_kzg_proof( | ||
| # blob_sidecar.blob, blob_sidecar.kzg_commitment, blob_sidecar.kzg_proof)`. |
There was a problem hiding this comment.
Double-checked with @mratsim , the proposer sig check is the faster one, so doing it before the KZG check: "pairings are 2 levels deep in KZG and 1 deep in BLS signature (just after deserialization and sanity checks that points are on curve and non-infinity)"
Batching of the two verifications would be clunky, likely not worth it.
| kzg_proofs: blobsBundle.proofs, | ||
| blobs: blobsBundle.blobs) |
There was a problem hiding this comment.
Sidecars are no longer sent back and forth, as they no longer need a signature. Instead, just the kzg_proofs and blobs are sent.
kzg_proofs and blobs are downloaded by the VC and then submitted again to the BN to support use case where the block is produced by a different BN than the one who is broadcasting. The BN then converts the SignedDenebBlockContents to SignedBeaconBlock + BlobSidecar list
| KzgProofs* = List[KzgProof, Limit MAX_BLOB_COMMITMENTS_PER_BLOCK] | ||
| Blobs* = List[Blob, Limit MAX_BLOB_COMMITMENTS_PER_BLOCK] | ||
| BlobRoots* = List[Eth2Digest, Limit MAX_BLOB_COMMITMENTS_PER_BLOCK] |
There was a problem hiding this comment.
Not sure where to put them and what to put as the correct limit. beacon-API uses 0 .. MAX_BLOBS_PER_BLOCK in informal metadata description, builder-API uses MAX_BLOB_COMMITMENTS_PER_BLOCK in a more explicit specification. In practice, it doesn't matter as the SSZ messages are not hash-tree-rooted in neither of the two protocols. If they were hash_tree_rooted, using the theoretical capacity (MAX_BLOB_COMMITMENTS_PER_BLOCK) makes more sense (that is also what they use for the KzgCommitments inside consensus, where HTR matters).
| @@ -69,25 +69,18 @@ type | |||
|
|
|||
| # https://github.com/ethereum/builder-specs/blob/534e4f81276b8346d785ed9aba12c4c74b927ec6/specs/deneb/builder.md#blindedblobsidecar | |||
| BlindedBlobSidecar* = object | |||
| block_root*: Eth2Digest | |||
There was a problem hiding this comment.
BlindedBlobSidecar may go away entirely, but the specs are still in flux:
For now, have done the minimal changes to it to make the flow work in theory, but the spec may optimize further and get rid of the BlindedBlobSidecar altogether.
|
This should be all that's needed to get us going with the new blob sidecar flavor, as an initial PR upgrading to v1.4.0-beta.4 test vectors.
Notably missing:
|
BlobSidecaris no longer signed, instead use Merkle proof to link blobs with block.Associated beacon-API / builder-specs still TBD; minimal changes done to compile in similar style to previous spec, but not standardized yet.