This is the third in a series analyzing the pettycoin implementation against Gregory Maxwell’s writeup on scaling. The first talked about UTXO commitments vs backrefs and the second talked about Propogation servers vs prev_txhashes.
In bitcoin, miners take any difference between the amounts input in a transaction, and the amount output; this is called the transaction fee. When designing a system where nodes don’t know every transaction, how do you verify that the miner has claimed the correct fee?
Every 2016 blocks, the hash of that block is used to select a random transaction in each of the previous 2016 blocks. The fee on this random transaction is multiplied by the number of transactions, giving the mining reward.
While miners will fight to get a block which gives them a good reward, this should cancel out as a block which gives one a good reward is likely not to do so for the others.
This solution involves hashing the fee amount in with the transaction to form the merkle tree; you prefix the transaction with the fee total and hash that, and also sum the totals for the two sub-nodes.
This lets you audit any transaction to ensure it’s been summed correctly. On the downside, it adds some bytes (say, 4) to each proof in the merkle hash (usually 32 bytes). This can replace the normal merkle hash, but that makes it unlike bitcoin and you might need both depending on how flexible sidechains are (assuming you want to be a sidechain).
The fee hashing is a better alternative: this way a miner knows what the reward for a block is. The alternative means that you might not get any reward if you include a transaction without a fee; though equally, you might get a huge one, psychology tells us that humans fear loss more than hope for gain.