whoever relays stuff 🇵🇸🏴☠️🇺🇲 on Nostr: Gaslighting about system resource usage is the fundamental problem in recent ...
Gaslighting about system resource usage is the fundamental problem in recent discussion on #nostr about spam filtering and the "web of trust" model.
People are getting suckered by the idea that we shouldn't have a computationally intensive enough process to filter spam properly, because instead nostr should be fast and the spam filter should suck? That's stupid. The discussion should be that vs not that. Instead it's just everyone arguing within the confines of... that.
Nostr's strength is supposed to be censorship resistance, not speed. Let X be fast and let X users not care if the spam filter can actually tell the difference between a new person and a bot. Let nostr be able to tell the difference between a new person and a bot.
---------
Below this line is nothing important, just an example of how spam filtering could work, which smarter people than me can figure out if they stop pussyfooting around
--------
Here's an example of how it should work in my view, but people will cry about it being supposedly impossible to optimize all this computation in a way relays could actually do reliably.
* Start with weighted, contextualized lists of trusted npubs
* Set some default values in user settings (adjust these as needed) ...
some examples that might be balanced enough (defaults might need to be changed, user definitely needs to be able to change all of these)
---- Minimum proof of work for posts to show up from outside web of trust: 0! None! Still user adjustable of course
---- 50 points of report weight filters a post from outside web of trust
---- Starting at 25 binary 0s, each additional binary 0 in proof of work cuts the weight of all reports in half
---- Following someone or adding them to your contact list cuts the weight of all reports on them by 100%
---- Contact list also makes reports FROM them weigh 100 points, and makes a user report weigh 50 points on all posts
---- Reports from someone followed by someone you follow - 50 points for a post, 25 points across all posts for a user
---- Reports from someone followed by someone followed by someone you follow - 25 points for a post, 10 points for all posts for a user
---- Reports from 4 follows away (final stage) - 10 points for a post, 5 points across all posts for a user
* When onboarding yourself, write proof of work notes to resist spam filtering until you have a strong web of trust presence, and then maybe still use web of trust for important or controversial notes or if you're highly targeted
* When onboarding friends, tell your client/relay to accept their npub so they start to have a web of trust presence
* When onboarding strangers, tell them to feel free to send you their npub to check out their posts or consider following them so they can get a web of trust presence started
People are getting suckered by the idea that we shouldn't have a computationally intensive enough process to filter spam properly, because instead nostr should be fast and the spam filter should suck? That's stupid. The discussion should be that vs not that. Instead it's just everyone arguing within the confines of... that.
Nostr's strength is supposed to be censorship resistance, not speed. Let X be fast and let X users not care if the spam filter can actually tell the difference between a new person and a bot. Let nostr be able to tell the difference between a new person and a bot.
---------
Below this line is nothing important, just an example of how spam filtering could work, which smarter people than me can figure out if they stop pussyfooting around
--------
Here's an example of how it should work in my view, but people will cry about it being supposedly impossible to optimize all this computation in a way relays could actually do reliably.
* Start with weighted, contextualized lists of trusted npubs
* Set some default values in user settings (adjust these as needed) ...
some examples that might be balanced enough (defaults might need to be changed, user definitely needs to be able to change all of these)
---- Minimum proof of work for posts to show up from outside web of trust: 0! None! Still user adjustable of course
---- 50 points of report weight filters a post from outside web of trust
---- Starting at 25 binary 0s, each additional binary 0 in proof of work cuts the weight of all reports in half
---- Following someone or adding them to your contact list cuts the weight of all reports on them by 100%
---- Contact list also makes reports FROM them weigh 100 points, and makes a user report weigh 50 points on all posts
---- Reports from someone followed by someone you follow - 50 points for a post, 25 points across all posts for a user
---- Reports from someone followed by someone followed by someone you follow - 25 points for a post, 10 points for all posts for a user
---- Reports from 4 follows away (final stage) - 10 points for a post, 5 points across all posts for a user
* When onboarding yourself, write proof of work notes to resist spam filtering until you have a strong web of trust presence, and then maybe still use web of trust for important or controversial notes or if you're highly targeted
* When onboarding friends, tell your client/relay to accept their npub so they start to have a web of trust presence
* When onboarding strangers, tell them to feel free to send you their npub to check out their posts or consider following them so they can get a web of trust presence started