2
One Bitcoin core mining policy is CPFP - child pays for parent. That is, transaction with high fee pays for its parent.
Does it also work the other way - that is, does transaction with high fee pays for child transaction that has low fees, given that they are both in the mempool? Or is the policy only "one way", from child to parent.
2
Does it also work the other way - that is, does transaction with high fee pays for child transaction that has low fees, given that they are both in the mempool? Or is the policy only "one way", from child to parent.
The policy is one-way: the parent doesn't pay for the child.
The child can't be included unless the parent is included, so effectively the miner can choose to include none of them, only the parent, or both the parent and the child. The miner will choose what pays the most for its resources. This means sometimes it will include a low fee parent in order to be able to include a high fee child but it only works in that direction.
1
One Bitcoin core mining policy is CPFP - child pays for parent. That is, transaction with high fee pays for its parent.
This works because a high paying child transaction will incentivise a miner to also include a low-paying parent transaction so as long as the transaction fee for the entire transaction chain averages out to be worth more on a fee/byte basis.
Does it also work the other way - that is, does transaction with high fee pays for child transaction that has low fees, given that they are both in the mempool? Or is the policy only "one way", from child to parent.
This is a given since a high fee transaction is naturally incentive-compatible and a miner will include it regardless even if there's a low paying child. The only consequence is that the child transaction may never be selected to be included in a block if the fee/byte is insufficient.
I am not sure if we understand each other. If there is a transaction A with a high fee and transaction B with a low fee, and B is child of A, and both are in mempool, are the fees averaged for the chain A-B and the parent help the child be confirmed sooner? – Karel Bílek – 2017-10-05T22:30:00.187
No. Since at the time that the parent is included, the child may not have existed. So the the policy only applies "one-way". On top of that, there's no reason to consider the child transaction anyways because there's no dependency on the child transaction to reap the Transaction Fee rewards. – renlord – 2017-10-05T22:51:35.150