IMHO regarding coverage: clients don't need to find a minimum set with maximum coverage. Clients can just connect to 500 relays to follow 500 people. It still works. But of course it is more efficient to do otherwise, and especially if some relays are down, to find the set that is up and covers everyone.
I think (as you describe in (8) that real world considerations probably far outweigh theoretical algorithm choices. But that doesn't mean we shouldn't bother to get the algorithms right.
Mike Dilger ☑️
npub1ac…9p35c
2026-02-26 20:30:28 CET
in reply to nevent1q…20xd
Author Public Key
npub1acg6thl5psv62405rljzkj8spesceyfz2c32udakc2ak0dmvfeyse9p35cPublished at
2026-02-26 20:30:28 CETEvent JSON
{
"id": "17f565a61ddaf982f1a15bc333c2218b905d911a1b37da029fd68d48feaa2b7e",
"pubkey": "ee11a5dff40c19a555f41fe42b48f00e618c91225622ae37b6c2bb67b76c4e49",
"created_at": 1772134228,
"kind": 1,
"tags": [
[
"p",
"17538dc2a62769d09443f18c37cbe358fab5bbf981173542aa7c5ff171ed77c4"
],
[
"p",
"9ec7a778167afb1d30c4833de9322da0c08ba71a69e1911d5578d3144bb56437"
],
[
"p",
"4a0510f26880d40e432f4865cb5714d9d3c200ca6ebb16b418ae6c555f574967"
],
[
"p",
"3bf0c63fcb93463407af97a5e5ee64fa883d107ef9e558472c4eb9aaaefa459d"
],
[
"p",
"e771af0b05c8e95fcdf6feb3500544d2fb1ccd384788e9f490bb3ee28e8ed66f"
],
[
"e",
"bcb29c08a3efe1c101858d0476b1c6d24690e3438978fd3893d3a0c9fc6224c8",
"wss://pyramid.fiatjaf.com/inbox",
"root",
"17538dc2a62769d09443f18c37cbe358fab5bbf981173542aa7c5ff171ed77c4"
],
[
"e",
"a11d74815647fc724e9d9785a005e26b06920a4405c82cbc013f61574a8718bb",
"wss://relay.damus.io/",
"reply",
"17538dc2a62769d09443f18c37cbe358fab5bbf981173542aa7c5ff171ed77c4"
]
],
"content": "IMHO regarding coverage: clients don't need to find a minimum set with maximum coverage. Clients can just connect to 500 relays to follow 500 people. It still works. But of course it is more efficient to do otherwise, and especially if some relays are down, to find the set that is up and covers everyone.\n\nI think (as you describe in (8) that real world considerations probably far outweigh theoretical algorithm choices. But that doesn't mean we shouldn't bother to get the algorithms right.",
"sig": "1cc4b49c694056aed16eb32fac6fe3e235495cfba13a626490d0f112cad670ba41dcef86c11d91c6d355d1df4fa793ff8cbb6ffff308ce9fd5a1e13a871e992d"
}