<gingeropolous> dayum: "All 2,210 hosts up" -- monerosim can do a full reproduction of that Nyx thing on the 1tb box. It turns out if your not slinging txs around we can cram quite a bit of monerods in there

<rucknium> Their reproduction instructions are giving me a syntax error. You are our only hope now :D

<rucknium> I think 1000 of those hosts are just py-levin, not actual nodes.

<gingeropolous> yeah i recall seeing the bot rambling on about creating a levin thing in python

<gingeropolous> it estimates another hour for the run to complete. can you access the lever65 stuff on senior? feel free to dig through the results / data

<gingeropolous> oh snap its done

<gingeropolous> "Interpretation — a clean, honest full-scale result. The established victim was gradually eclipsed from 0 to a

<gingeropolous> ~7/12 plateau (peak 8/12), shedding every seed/miner connection — but it did not reach a clean 12/12. The

<gingeropolous> residual 4–7 slots stayed benign, held by the paper-faithful 1,199-node honest population re-injecting records

<gingeropolous> faster than update_sync_search evicts them. This is exactly the convergence-limited behavior we predicted,

<gingeropolous> and it quantifies the honest-population dependence: fewer benign → higher takeover (585-node Nyx hit 9/12; the[... more lines follow, see mrelay.p2pool.observer/e/z9Hz_KoLRy05UUJK ]

<rucknium> Thanks! So it doesn't reproduce the results. I don't recall reading anything about "whistler-filling" in the paper. Try to ask it to reproduce the behavior in the paper. I guess you already did that, but why does it say "we did something besides what the paper said"?

<gingeropolous> yeah i dunno. prolly cause its opus 4.9

<gingeropolous> 4.8

<gingeropolous> ah ok. apparently it had its sights set on seeing if we could run that scale at all.

<gingeropolous> i guess that means another 20 hour run....

<gingeropolous> pastebin.com/raw/2TqvFGCd for its rationale

<gingeropolous> crankin it now

<gingeropolous> "The paper's attack (formally N-I/N-II/N-III "network-wide peerlist poisoning", not "whistler-filling" — my mislabel)" >> this guy