docs(protocol): add comment for EIP-1559 (#16389)

Co-authored-by: Brecht Devos <[email protected]>

packages/protocol/contracts/L2/Lib1559Math.sol

@@ -5,7 +5,8 @@ import "../thirdparty/solmate/LibFixedPointMath.sol";
 
 /// @title Lib1559Math
 /// @notice Implements e^(x) based bonding curve for EIP-1559
-/// @dev See https://ethresear.ch/t/make-eip-1559-more-like-an-amm-curve/9082
+/// @dev See https://ethresear.ch/t/make-eip-1559-more-like-an-amm-curve/9082 but some minor
+/// difference as stated in docs/eip1559_on_l2.md.
 /// @custom:security-contact [email protected]
 library Lib1559Math {
     error EIP1559_INVALID_PARAMS();

packages/protocol/docs/eip1559_on_l2.md

@@ -33,3 +33,9 @@ It turns out the initial value of gasExcess doesn't really matter for the above
 ## Adjust the slope
 
 To adjust the slope of the curve to satisfy $R == basefee(2T)/basefee(T)$, we simply need to choose $M$ and $b_0$. $b_0$ is simply to decide -- if we believe the cost of a L2 transaction is $1/n$ of the same L1 transaction, we simply use the current L1 base fee divided by $n$. Then we can simply tune $M$ to make sure $R == basefee(2T)/basefee(T)$ holds. This is very simply manually a try-and-adjust approach as shown in `Lib1559Math.t.sol`. The TaikoL1 contract will check if $R == basefee(2T)/basefee(T)$ holds but will not calculate $M$ for us.
+
+## Implementation Difference
+
+Our implementation computes the basefee prior to applying the gas used by the current block (indicated by the blue line in the image below), in contrast to Vitalik's concept, which calculates the average basefee by also taking into account the gas consumption of the current block (represented by the red line).
+
+![Merkle Proof](./images/1559.png)