09:07 <Sjors[m]> "rogue calls in mining functions" - ominous

14:06 <bitcoin-git> [bitcoin] krishpranav opened pull request #31735: Fix compilation errors in blockchain.cpp related to serialization, UTXO dumping, and path conversion (master...master) https://github.com/bitcoin/bitcoin/pull/31735

14:15 <sipa> glozow: a random idea about orphan memory accounting, to get the best of both worlds (guaranteed per-peer orphan space, but higher global limit while not under attack): every peer has an orphan reservation size (e.g. 4MW per outbound, 400kW per inbound), and the global orphan limit is the sum of the reservations (which is thus dynamic). Orphans are accounted for for *each* peer (regardless of

14:15 <sipa> however many peers announced jt). Orphan evictions only happen...

14:15 <sipa> when the global limit is exceeded, and find the peer whose usage/reservatiom ratio is the highest (and only removes the orphan from that peer, others may remain, and thus more evictiond may be needed).

14:16 <sipa> as long as no peer goes above their reservation, no evictions can happen, so it is at least as good in that regard as the alternative of just having strict per-peer accounting

14:18 <sipa> I don't know if any of this is necessary though, I think strict per-peer accounting is fine, and is more categorically safe in the sense that attackers cannot even influence eviction decisions for other peers