Leo Wandersleb on Nostr: Trying to understand this ... So you have a list of relays and each has a list of ...
Trying to understand this ...
So you have a list of relays and each has a list of inbox pubkeys in a probabilistic filter and when you publish an event, you publish to the author's outbox relays and check every other known relay for being the inbox of any of the mentioned pubkeys via the probabilistic filters? And that lookup delays sending by 255ns at the cost of 10MB in RAM? Or is it 255ns per user or per user per relay? I guess the latter? So if you have 100 relays indexed as being the inbox relays of pubkeys and you have an event mentioning 40 pubkeys, you spend 40 x 100 x 255ns = 1ms to compile the list of relays to publish the event to?
So you have a list of relays and each has a list of inbox pubkeys in a probabilistic filter and when you publish an event, you publish to the author's outbox relays and check every other known relay for being the inbox of any of the mentioned pubkeys via the probabilistic filters? And that lookup delays sending by 255ns at the cost of 10MB in RAM? Or is it 255ns per user or per user per relay? I guess the latter? So if you have 100 relays indexed as being the inbox relays of pubkeys and you have an event mentioning 40 pubkeys, you spend 40 x 100 x 255ns = 1ms to compile the list of relays to publish the event to?