I really like Ryan Fugger's original vision of a generalized peer-to-peer money system, where "customers" are also banks, and the idea to run such trustless as well with payment channels.
But, a simpler variant, the full mesh banking network -
https://meshnet.grok.me - has a few benefits:
- Path-finding disappears - there's a clear social geography whereas in web-of-trust there is not.
- At most 3 hops means payment coordination can be simplified, "staggered deposits" rather than "staggered timeouts" (no chained timeouts)
The bests architecture I can see has a HTLC only on the last hop - thus no chained timeouts. It's between a bank and a customer, the timeout can be as long as there is trust to not attack it (i.e., liquidity attack by stalling). With no trust, probably pretty short timeouts... with trust or a customer proven to not liquidity attack by their history, can do longer timeouts.
So, "HDLC", hash-deposit locked contract, with "staggered deposits" on hop 1 and 2, and HTLC on hop 3.
Of course, "staggered deposits" becomes ridiculous in generalized multihop as it is 2^N-1, exponential, but for 3 hops (and since HTLC can be done on last hop, without any chained timeout), it's just 2^2-1 = 3.
No chained timeouts means the HTLC does not mean an intermediary risks getting stuck having paid out without getting paid from upstreams. It only means the payment never started. This is a big difference... That's why timeout on last hop does not have the drawbacks of chained timeouts...
Another benefit of this over LN is I assume network will have more liquidity - whereas a typical web-of-trust like distribution has quite low (of course, LN can generalize into forming these types of structures too).
The drawback of this is it's centralization of power. But, the benefit over the historical banking full mesh, is the banks relate to one another trustlessly. So it's still a step forward.
Note, I perfected general multihop last year (the 3-phase commit, see
https://resilience.me/3phase.pdf), and I am in no way against general multihop. I just think sometimes a "dumber" system can be good too. Fewer parts. Maybe it's not one or the other...
With this full mesh banking network, any crypto-currency would reach all capacity it would need... While still being trustless, but, the routers do acquire a slight monopoly, but can be bypassed by sending tx on-chain, or by opting-in to another routing network, or with something like LN. So, they might behave out of their interest in getting the routing fees and not exploit position much, it's an improvement over the current system (which is pretty much this but trustbased, oligarch international bankers that trust one another internally and hate the rest of the world).
