Signing a transaction: What is actually signed? [noob question]

Question

I'm trying to wrap my head around how signing transactions actually works. As I understand ECDSA signatures roughly work like this: There is a magic function, let's call it sign that takes two parameters. The private key and the message. It outputs the signature or in the case of web3.py or web3.js, three weird values v, r, and s.

So when users sign a message they are basically signing a blob of data which is the message value in the sign function. Am I correct in this assumption? So when users sign a message they are basically signing a blob of data which is the message value in the sign function. Am I correct in this assumption?

How can I access that data blob? Is it possible to get it from a public transaction via web3?

I found this article and there it says that the data hash that will be signed is created from rlp + hash. Has anyone an idea what rpl means or how to get the output hash with web3?

Answer 1

ECDSA works with numbers so to sign a message first you have to encode it as an number.

The precise meaning of RLP is defined in the Yellow Paper (Appendix B Recursive Length Prefix). It is a function that will encode something structured (a transaction for example) as a sequence of bytes. Given a transaction Transaction then RLP(Transaction) returns a sequence of bytes.

Actually ECDSA numbers they also have to be in a range. To ensure the RLP(Transaction) is withing the correct ECDSA range we apply a hash function like KECCAK256(RLP(Transaction)).

Also the exact details of the ECDSA signature are in the Yellow Paper (Appending F Signing Transactions).

r, s, v = ECDSASign(KECCAK256(RLP(Transaction)), PrivateKey)