isFullyCanonicalSig.js

GROUP_ORDER = 'FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFEBAAEDCE6AF48A03BBFD25E8CD0364141'
// Divide by two and round up = shift right 1 bit and add 1.
HALF_GROUP_ORDER = '7FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF5D576E7357A4501DDFE92F46681B20A1'

function isFullyCanonical(signatureDer) {
  s = signatureDer.slice(-32)
  // A lexicographical compare between big-endian hex strings yields the same ordering
  // as a comparision of the unsigned integers they represent.
  return s < HALF_GROUP_ORDER
}
  
signatureDer = '30440220...'
isFullyCanonical(signatureDer)

signatureDer is just hex string of the sig?

To be honest, it depends on how the signature is represented... The signature is two 32- or 33-byte numbers commonly named r and s. The normal DER encoding includes both of them, but there is a compressed encoding that takes only one byte from one of them, but I forget which. To judge full canonicalization we need the full s. Do you have an example from the database?

I'm remembering wrong. The public key is compressed, but the signature is not. So yes, signatureDer is just the hex string of the signature. There are multiple representations for the signature, but the one we use in XRPL is DER-encoded.

Txn with hash 58F140BD9B4FB4160F4C2B3711F06E24F46C7085F8CF9A6FE119F93AEAC7B697 has signing public key
03D847C2DBED3ABF0453F71DCD7641989136277218DF516AD49519C9693F32727E and signature
304402207374724663A13CDF92CFAA9BAEFC085D3379E60B323DA445506398B19299ABE002203CD9826F711D0F378171B35105E79F4A346BBC8208402B6F866E509CD453F4F3

All these are bigquery SQL STRING types.

Yes. Let me break down the DER encoding of the signature:

• 3044: sequence of (44 in base 16) = 68 bytes
• 0220: integer of (20 in base 16) = 32 bytes (if a positive integer has 1 for the most-significant bit of its most-significant byte, it is prefixed with a zero byte 00)
• ... 32 bytes of r ...
• 0220: integer of (20 in base 16) = 32 bytes
• ... 32 bytes of s ...

Just so I understand -- for this example, s is after 0220. s in this case is 3CD9...F4F3, and a canonical sig because less than HALF_GROUP_ORDER? Should it be signatureDer.slice(-64) instead?