Three new projects were released recently - Wayman, Filestr and Zapworthy, all relying on new data types (kinds). So we now have music, files and highlights data formats (nostr events), all queryable over a common API (nostr protocol). I don't know who needs to hear this, maybe it's just me, but I finally realized it's the true potential of Nostr.
We used to just have "information" on the internet - a presentation-layer (html+css) and custom APIs behind it on every website. Now with Nostr we can access "data" - all data types can have a standardized representation as nostr events and a standardized API to store/query this data from a decentralized swarm of relays.
We already have people, notes, posts, highlights, stalls, products, music, files. All the other stuff will follow - companies, drugs, recipes, cars, TVs, games, excel tables, you name it... All visible to any new app, all addressable, composable into new experiences.
I thought that portable identity was what Nostr added to the internet. Now I realized it's also a data layer.
This is mind blowing! Need to rethink it all over again.
> There seems to be a misallocation of resources in a misguided desire to push various data types through a text protocol.
I don't like the idea of storing 'files or other stuff' on Nostr. You only need metadata. Just like an aggregator doesn't store the songs/tickets/goods itself, but only links to them.
It's not about storing everything on Nostr, it's about indexing everything on Nostr.
Does this point change anything in your line of thinking?
Another spam-free-relay experiment:
wss://relay.nostr.band/trusted
This endpoint (with /trusted path) has all the nostr data from all relays, but it hides all posts and profiles that have low trust rank, unless you specifically ask for them by id or pubkey.
In case of a spam attack on the network you can remove 'read' permissions from all other relays (or click 'disconnect' if your client allows that), and leave 'read' for this relay, and have content from all the network visible, except for spam. When attack is over - re-activate read permissions or re-connect to other relays.
Trust rank is an internal metric similar to PageRank. The upside of it is that it's very robust against sybil attacks, i.e. bots that interact with themselves to pump their likes/follows/vanity metrics gain nothing since only interactions from broader network would count.
The downside of using trust rank for spam-filtering is that new blank accounts that haven't had any interaction from trusted accounts will not be visible, so you won't be able to see and greet new nostr users.
The relay is free for now. Please let me know what you think of this approach.
Update on https://nostr.band:
- view profile edit history - important!
- view relays that store a profile or a post
- better UX
For more info on why profile edit history might be useful, check this thread #[4]
A couple weeks ago someone was sending Hello Nostr from new empty accounts every second, so global feed was useless. Every message contained some POW even. It wasn't really malicious though - it didn't try to send to every single relay, didn't change the content, and didn't send 100x as much. So a couple relays were banned by clients, and "Hello Nostr :)" was blacklisted, and then the spammer turned it off.
Then there was a guy who was thrashing another one by answering to random threads from several accounts, and no one had any good idea of how to fight that.
The only reason you don't see that every single day is because no one is doing it atm. It's just a question of time until nostr annoys someone and it comes back.
When people talk about spam, they generally mean 2 things: high-volume cheap spam from lots of empty accounts, or a low-volume aggressive marketing from well funded non-empty accounts. The first one (it's DoS essentially) can only be properly prevented by LN paywall. The second one (let's just call it 'noise') - can be prevented by reputation filters, of which I believe TrustRank to be the proper one.
We could use TrustRank even now to filter the first type of spam (I do at nostr.band), but then new people joining nostr would not become visible in a global feed without getting some friends to follow and interact with them (without gaining some reputation). LN is better because it filters out high-volume spammers without hurting the normal noobs.
Agreed. Although non-canonical kinds is just people experimenting with other uses of the protocol - irrelevant kinds don't influence the popular nostr clients focused on kind:1 posts. It might be a problem for relays, but then they could just ban useless kinds, so it's not a big issue.
I'm working on an 'algorithmic feed' problem. You can check out https://nostr.band homepage for trending urls, hashtags and people. Trending posts coming soon. These lists are free of several kinds of spam that I've seen on nostr up until now (not based on simple easily-inflatable like/reply/follower counts). Is this something you're looking for? What other kinds of 'tops' would you consider useful?
Thank you for your detailed perspective.
I believe a general robust 'show me popular content' solution is TrustRank, similar to PageRank that Google successfully uses for the decades-old open ecosystem, which is Web.
If you add your following list, then you could get 'popular content people I follow engaged with', which is kinda what Twitter does.
With regards to a genuine question being downvoted by the crowd - that seems hard, need to think on that.
A couple weeks ago someone was sending Hello Nostr from new empty accounts every second, so global feed was useless. Every message contained some POW even. It wasn't really malicious though - it didn't try to send to every single relay, didn't change the content, and didn't send 100x as much. So a couple relays were banned by clients, and "Hello Nostr :)" was blacklisted, and then the spammer turned it off.
Then there was a guy who was thrashing another one by answering to random threads from several accounts, and no one had any good idea of how to fight that.
The only reason you don't see that every single day is because no one is doing it atm. It's just a question of time until nostr annoys someone and it comes back.
That's right. OTOH, I'm thinking about Alby and stacker.news giving out free NIP names might get abused by spammers who will join just to get a good looking name. Paid names are a better approach IMHO. Time will tell though
Well it's really not an identity, but an alias, a pseudonym. A domain owner or DNS could a) steal your alias or b) erase it. In case b) you just set another alias, but case a) might be a bigger problem - people could try to impersonate you and scam others.
I've seen people in Nostr design telegram group proposing visual markers on avatars that depend on pubkeys so that users would notice a different pubkey impersonating a familiar one. Haven't seen that implemented anywhere yet though
Notes by brugeman | export