The secp256k1 curve permits the computation of the public key of signed digest when coupled with a signature, which is used implicitly to establish the origin of a transaction from an Externally Owned Account as well as on-chain in EVM contracts for example, in meta-transactions and multi-sig contracts.
Currently signatures require 65 bytes to represent, which when aligned to 256-bit words, requires 96 bytes (with 31 zero bytes injected). The yParity in RLP-encoded transactions also require (on average) 1.5 bytes. With compact signatures, this can be reduced to 64 bytes, which remains 64 bytes when word-aligned, and in the case of RLP-encoded transactions saves the 1.5 bytes required for the yParity.
The motivations for a compact representation are to simplify handling transactions in client code, reduce gas costs and reduce transaction sizes.
A secp256k1 signature is made up of 3 parameters, r, s and yParity.
The r represents the x component on the curve (from which the y can be
computed), and the s represents the challenge solution for signing by a
private key. Due to the symmetric nature of an elliptic curve, a yParity
is required, which indicates which of the 2 possible solutions was intended,
by indicating its parity (odd-ness).
Two key observations are required to create a compact representation.
First, the yParity parameter is always either 0 or 1 (canonically the values
used have historically been 27 and 28, as these values didn't collide with other
binary prefixes used in Bitcoin).
Second, the top bit of the s parameters is always 0, due to the use of
canonical signatures which flip the solution parity to prevent negative values,
which was introduced as a constraint in Homestead.
So, we can hijack the top bit in the s parameter to store the value of
yParity, resulting in:
The compact representation proposed is simple to both compose and decompose in clients and in Solidity, so that it can be easily (and intuitively) supported, while reducing transaction sizes and gas costs.
The Compact Representation does not collide with canonical signature as it uses 2 parameters (r, yParityAndS) and is 64 bytes long while canonical signatures involve 3 separate parameters (r, s, yParity) and are 65 bytes long.
The ethers.js library supports this in v5
as an unofficial property of split signatures (i.e. sig._vs), but should be
considered an internal property that may change at discretion of the community
and any changes to this EIP.
There are no additional security concerns introduced by this EIP.
Copyright and related rights waived via CC0.