here are my thoughts from what I have read from your posts:
nostr is inherently an unreliable transmission medium, as notes may be rejected, reordered, delayed, or not delivered for a multitude of reasons and relays will be built with this assumption
the current design doesn’t implement any flow control or retransmission I believe, so this can be a problem
implementing another TCP like layer on top of nostr events is vulnerable to a meltdown, where the lower layer having losses causes delays in the upper layer causing retransmissions and more load on the lower layer
there is also the fact that nostr and IP have different service quality goals
on nostr latency can be in the seconds, caching will exist and QoS can be applied for different events
IP has some loss and reordering tolerance, but it is expected that paths have mostly constant latency and throughput, and that path switches happen rarely
/ semisol
npub122…cgrkj
2024-07-18 10:41:07
in reply to nevent1q…uecg
Author Public Key
npub12262qa4uhw7u8gdwlgmntqtv7aye8vdcmvszkqwgs0zchel6mz7s6cgrkjPublished at
2024-07-18 10:41:07Event JSON
{
"id": "7d02a7cf598eaab310a38fcc19e5176a384df5e5b23669b460c9104be6d0e963",
"pubkey": "52b4a076bcbbbdc3a1aefa3735816cf74993b1b8db202b01c883c58be7fad8bd",
"created_at": 1721292067,
"kind": 1,
"tags": [
[
"e",
"1e2864347f577bc8a61fedefd3537a1d3a1c929f9b02168a1b1ecba940eead33",
"wss://nostr-pub.wellorder.net",
"root"
],
[
"e",
"d9a1a30723779704ae3cb822fa418c8bfa6a7d2af17b6e2dac4747728ae996a4",
"",
"reply"
],
[
"p",
"50d94fc2d8580c682b071a542f8b1e31a200b0508bab95a33bef0855df281d63"
],
[
"p",
"8cda1daafb2df8daf4500533cba69b645afa0caf7136e341be1b66a2e9e5bdd8"
]
],
"content": "here are my thoughts from what I have read from your posts:\n\nnostr is inherently an unreliable transmission medium, as notes may be rejected, reordered, delayed, or not delivered for a multitude of reasons and relays will be built with this assumption\n\nthe current design doesn’t implement any flow control or retransmission I believe, so this can be a problem\n\nimplementing another TCP like layer on top of nostr events is vulnerable to a meltdown, where the lower layer having losses causes delays in the upper layer causing retransmissions and more load on the lower layer\n\nthere is also the fact that nostr and IP have different service quality goals\n\non nostr latency can be in the seconds, caching will exist and QoS can be applied for different events\n\nIP has some loss and reordering tolerance, but it is expected that paths have mostly constant latency and throughput, and that path switches happen rarely",
"sig": "ecbb65bce48d5302df3251c834d02372a041be67de53e81eb851e08c7146c885e717bbbb684dfa3c1a54502a1012c633f9cdd475749d153e3c2c6b03d1e3ae23"
}