test if routing hints from erring nodes are useful

[renepickhardt]

Rusty mentioned yesterday that he considers giving an order of magnitue of the balance in a failing channel inside the error onion.

I just realized if a node has not enough balance to forward an onion to its peer P instead of giving a hint of what could be forwarded on this channel to P in the error it would be rather useful to send back a list of channels that could forward that amount (especially if the peers of those channels are not too far away from the peer P to which the payment should have been forwarded originally)

I think once we start our experiments it should be easy to stimulate this and to measure the effect. My feeling is that the boost should be quite big as we started on a likeli path (everything else would be worse) we already have certainty to reach a node and we would get certainty about the neighbors if that node gives us routing hints.

In case this turns out to be useful I think the peers of P thing should be rather trivial on the public network. My node can have a distance map for each of my peers since the network topology does not change often or could be updated with every block. I would just compute the shortest paths between all my peers (with me removed from the network) and for every peer maintain a sorted list. Anyway I guess it will take quite some time until we have results. But looking forward to include that experiment. Note that Rusty also mentioned randomization. of course if we conduct such an experiment we could also send back random but liquid channels