Anatomy of a Nostr DM spammer.
67K DMs since the 28th. No other events.
Happiest 24 hours NostrGraph has seen in a while.
Main spike was 2,500 DMs from WoS relay (doubled by now.lol, who aggregates). Minor new identity suspected spam past few hours too (bottom middle).
You can see the clear wave in timezone activity too 🙂
Otherwise a pretty happy dashboard. 
I believe Damus does now. New load at the bottom. Unsure if additional logic however.
I prefer it too.
I’m always interested in seeing new implementations.
Would an aggregation service for reactions and reply counts, etc help? Could reduce volume anyway.
By volume, reactions are typically highest.
No relays do it yet, however it could make sense to start returning metadata like my REST endpoint.
Still looking for a simple lightning payment API that can keep account of a pubkey’s credit and debt during API requests.
I’ve refactored the Nostr Proof of Work Service Provider codebase, in prep for open sourcing it this week.
I’d release it now, but I’d like some more testing with so much code change.
Thanks. I’ve seen these. I’m thinking for all commands, as I expect the list will grow too.
What I find interesting is how relays are tending toward common Webserver patterns. Kind of makes sense, but lots of other protocols didn’t.
What’s most interesting however, is with that kind of pattern, it accelerates innovation and modularity. Could be a key architecture change that drives the next generation of Nostr relays.
Is it an open source relay implementation?
Any Nostr relay’s architected around a middleware pattern?
Basically new commands or extensions could just be new middleware that’s optionally enabled.
Useful for things like whitelists, spam filtering, relay memberships, new relay commands (like AUTH or POW), caching, etc.
The key issue is it’s only captured at a point in time. The next time you check it could be different.
And who has been checked is also a subset of the entire network. Never full visibility.
It’s still valuable, however it’s decaying data.. the longer ago it was seen the less likely it’s still there.
There are worse ways to fail 🪺 🕶️
Around 700MB worth of valid (spam filtered) and uniqued events past 24 hours.
Some are ephemeral 20k kind, so not persisted.
And for clarity. That spam number isn’t de-duplicated. So it’s higher than a normal relay would see.
But it does show the challenge aggregators have, as it starts to scale pretty quickly and become a management challenge.
And currently around 6,000 spam events per minute. Or 100/second being filtered.
This is why POW and other methods to make spam more expensive are important.
Without spam filtering we would have 8.5MM new events per day. I only have 12.5MM unique events (spam filtered) today total.
NostrGraph currently processes 8,000 events/min (133/sec), across about 125 relays.
Around 1,500/sec are valid (+ spam filtered, but pre de-duplication). It varies through out the day.
Of those, around 150/min are unique events. Around 50-120/min of those are from unique pubkeys.
Around 220-250k unique events a day. Around 10k (named) unique pubkeys posting an event. Average 23/events per pubkey a day.
My NIP validation worker needs attention, but I’ve seen it as high as 23k.
I haven’t tracked lud06 or zaps seriously at present. Too many unverifiable.
Numbers are decent, but nothing has 100% visibility.