20 sats \ 4 replies \ @supertestnet 7 Jun \ parent \ on: Lightning payment fee preview lightning
the lightning network does not currently penalize nodes that send failed payments and for that reason payment probing is very common. The vast majority of payment attempts on lightning fail, largely because so many of them are not "real" payments, just probes. End users do not see these as failures because their wallets don't show them, since they are just "checking" to see if a future "real" payment (to be made seconds later) is likely to go through.
Lightning needs the ability to charge a non-zero fee for the right to hold an HTLC open. Channels don't always need to charge that fee. But we definitely need the ability to charge those fees when under attack.
Of course, this is ugly for wallets as failed payments will still cost money. But it's better than the alternative.
reply
Will nodes still be able to fail payments for any reason? If so, that implies that routing nodes can fail payments for whimsy's sake and still charge for them. If (1) routing nodes can fail any payment (2) still get paid for it and (3) reliably get other routing nodes to try to route payments through them, then this downside follows: scammers will spin up routing nodes, fail all payments, and watch the proceeds come rolling in.
Thankfully, routing nodes who regularly fail payments currently get quickly blacklisted by other routing nodes, so I don't think number 3 will be true. Trolls won't make much money once they're blacklisted. Assuming channel opening fees cost more than trolls can earn by trolling then the fact that they *can* do it shouldn't cause any serious problems, because there's a monetary disincentive to hopefully stop them from doing it. Here's hoping.
reply
Will nodes still be able to fail payments for any reason?
Of course. It's impossible to prove to a third party the reason why most failures happen.
But as you say, reputation and setup costs will limit the usefulness of that fraud. Channels have a pretty big cost to create, especially in the current system where you have in fact proved ownership of the txout.
reply
I always monitor failed htlcs going through my node and never see the reason that preimage was incorrect. So what you describe is still theoretical.
reply