Oddbean

Currently reading the Gossip v 1.75 proposals by Elle Mouton on delving/github (pr 1059 on lightning/bolts). In the PR there's a lot of discussion about gossip v2 and 'overcommit' but I've failed to much other than the original 2019 note from @Rusty Russell on gossipv2 and also https://lists.linuxfoundation.org/pipermail/lightning-dev/2023-July/004014.html which is some meeting notes. Is there anything else I can read to get up to speed on these ideas about channel/node announcements? Related: has LN mailing list "gone to delving for now", but not really (i.e. it's mostly just github)? Or? (Guess I'll ping @niftynei and Matt Corallo .. uh if i could find his nick!)

Private gossip efforts kinda stalled without a champion, sadly. Gossip v1.5 is mostly redoing gossip so that we can do taproot channel announcements and also minisketch the gossip sync protocol. It also theoretically supports announcing more aggregate channel capacity than UTXO proofs provided but unclear whether people will implement that (especially on the announcing side). Ideally someone with a bit of time could sit down and spec out (and then implement) log-scale ringsig announcements, but I’m not sure it’s gonna happen soon on its own.

Yeah I continue to find myself torn on the topic as I read people's thoughts about it. I have this visceral distrust in reputation systems, but that doesn't mean I don't see why people find them a natural choice. Distinction between attack behaviour and honest behaviour always feels like trying to solve the wrong problem, but I get that it just doesn't feel palatable to say "here's this really big cost to participate because i don't know if you're attacking and i'm deliberately not trying to make a choice". There was a long chat about this on the gist where I first wrote "RIDDLE" with a guy who proposed Chaumian tokens, and we kind of agreed on "prove ownership of scare resources with ZKP and use that to get Chaumian tokens (think privacypass) which are then convenient/performant to consume with good privacy". I mean if either privacy is not #1 priority, or you don't mind that bootstrap is not solved, then maybe just pay for the tokens. It might seem a bit high resource requirement if we tokenize access to routing nodes but I feel like .. maybe not? It doesn't have the specialization weakness of PoW and i don't think it's terribly costly. We still need it to cost what the resource is actually worth for the system to be densible but it's still a cost you can fine tune enough I think.