18:25:19 <UkoeHB> What do you guys think about enforcing fees with 2 significant digits in seraphis (the fee PR #7819 only makes it a convention)? This removes a convention (room for fingerprinting) and lets fees always be 2 bytes via scientific notation (conveniently, it can even be serialized as a varint to 2 bytes since the msb of both bytes will be unused).
18:26:38 <UkoeHB> ArticMine: ^
18:29:45 <moneromooo> The fewer significant digits, the more arbitrary the priority of a given tx compared to other txes in the pool, since miners will use fee/byte as a yardstick.
18:30:22 <moneromooo> I don't mean that as an argument against really, it might be something really minor.
18:31:52 <moneromooo> Thinking about it, it might mean that in times of pressure, a tx that scores low on fee/byte due to the quantization effect would always be left to wait, regardless of its age.
18:39:52 <UkoeHB> Transactions with low fee/byte will be always be left to wait. Quantizing fees will just rearrange the marginal txs at the bottom slightly.
18:56:17 <ArticMine[m]> I do see an issue with this
18:57:07 <ArticMine[m]> Sorry do NOT see an issue with this
18:59:28 <ArticMine[m]> The only thing in consensus  here is the number of significant figures, not the amount, so no problem
19:48:05 <sech1> what prevents us from making fee values a part of consensus? fee per byte is already calculated on the node and is a part of consensus, so tx fee can just store a multiplicator of the default fee - 1 byte should be enough for this
19:50:55 <UkoeHB> Fees are ddos protection for the network. I think that's why minimum fees are left to the discretion of nodes (if the algorithm is bad and a hard fork isn't available, nodes can change min fees).
19:55:25 <moneromooo> The fee depends on the previous block sizes. Since a tx isn't mined instantaneously, it'd change while the tx is waiting in the txpool, which it cannot since it'd cause the outputs' values to change, and thus break the signatures.
19:55:27 <sech1> hmm, I always though tx fees were checked when checking blocks but nope. They're only checked when adding tx to mempool
19:55:55 <UkoeHB> moneromooo: good point
19:55:58 <sech1> So some miner could mine transactions with literally zero fee? For example pool mining its own payouts
19:56:10 <UkoeHB> yes
22:26:57 <ArticMine[m]> One major problem with making fees part of consensus is that this encourages miners to collect or refund fees via a website. Very bad for privacy
22:27:29 <ArticMine[m]> There are many more issues with this