Open palango opened 5 years ago
I think this would be helpful only in combination with a "decay confidence" measurement from PFS: For each returned route, PFS should include a "best before" date:
PFS: with 95% confidence, given my knowledge about the network's past, this route will still work
x
seconds/minutes/hours from now.
Since Raiden aims at sub-second transfer times, any client side configuration setting for a fixed route invalidation interval (larger than few seconds) would be indistinguishable from a "best effort" approach (i.e. just exhaust all initially inquired routes). This is, because any node in a route can change its capacity in the blink of an eye.
See https://github.com/raiden-network/raiden/blob/develop/raiden/transfer/state.py#L183
A
RouteState
should have an associated timestamp, so the initiator can decide when invalidate routes and request new ones from the PFS.