A base fee of 0 makes it easier for algorithms to reason about the costs, as the cost function is linear.
A fee rate of 1ppm is extremely low, and as such channels with a fee rate of 1ppm attract lots of routing attempts. However, based on my experience, such channels are quickly drained and rarely serve a benefit to the network. As such, they attract lots of routing attempts that will fail, which also makes it harder to find cheap channels that actually work at the configured fee rate.
Furthermore, an inexperienced node operator using the default fee rate for a new channel might not be able to increase the fee rate before the liquidity is routed at 1ppm. This happened to me more than once, and a higher default fee rate would have helped in these situations.
Why not do it and see what happens? Is your point to get everyone else to do it too? I like the idea the default is high to avoid drainage before you lower it later, but 10kppm is quite high (1%) and renders the network mostly useless for any multi-hop transfer in excess of 50k sats ($15).
On another note, base fees should always be zero unless you're trying to traffic limit your channels for some reason.
reply
The point of getting it as the default in lnd (don't know about the other implementation's default) for all is that currently there are many fully depleted channels that hamper the reliability of the network. By putting a high default, one has to put some thought/effort into lowering it, thus decreasing the natural pool of depleted channels.
Now I wonder if 10,000 ppm basically blocking flow will then just lead to routing attempts to fail the other way. He probably addresses that somewhere in his links. Will check.
reply
I mean it will render the network useless because the fees will exceed a tolerable level. Is it reasonable to pay $2 to transfer $100? One of the saleable points of lightning was to allow fast AND CHEAP transfers of bitcoin balances. By raising the cost, you're only reducing usage of the network and not necessarily improving it. Sure more transactions will succeed theoretically, but render the network useless and you lose the network effect and you get less liquidity not more.
reply
The point is not to keep it at that value. It's just to start high and bring the values down until payments flow, rather than start very low and have it depleted before you can even raise it to a normal value. FYI, normal values are between 5 and 300 ppm for most of my channels, so not at all partisan of high static fees.
reply
That makes sense.
reply
This sounds reasonable but i wonder why are the fee rates configurable in the first place? Why isn't the base fee and ppm set to a static number for all nodes ? Seems like the constantly changing rates are the cause for most issues and what is the benefit? Genuinely curious.
Also im surprised ZFR is supporting this because he seems like the kind of guy that is monitoring the network for 1ppm channels and then taking advantage of them. No offence.
reply
I follow Carsten's logic and would probably support it if it becomes the standard. I have actually already changed the default values in the lnd.conf file so that indeed, I avoid having a channel drained before I've had time to adjust the fees by hand.
I don't think a static number for all nodes makes sense. I use dynamic fees to make sure all channels can work at their best (and, let's be honest, maximize profitability). With fixed fees, some channels would not route at all or would get depleted in no time.
ZFR does not do any active rebalancing of his channels if I remember correctly. So there is no reason for him to monitor 1ppm channels. He just lets things flow as rebalancing would incur a cost that he cannot recoup by routing fees.
reply