R
The problem with social-graph based trust indicators is that people are SO SLOPPY about who they follow. The plethora of followback parties, follow-for-follow, follow trains, follow-farming hashtags, etc, shows that-- I've been pleading with people for weeks to stop doing it and omg, the abuse I get
24,848,050 views
R
Replying to
@rahaeli
My new party trick is looking through the last 100-200 follows of anyone who huffily says "but *I* vet everyone I follow" and telling them how many I reported as scammers, spammers, and obvious impersonators
13,250 views
R
Replying to
@rahaeli
(It's always at least a double digit percentage, sigh.) I think the ideal solution is not going to be any single centralized "this person is really them" verification system but a mix of options, it'll just take a bit to figure out what the best options are for the mix.
13,250 views
R
Replying to
@rahaeli
I almost wonder if the best solution would be a specialized class of labelers (not interchangeable with moderation labelers) that can each set their own criteria and users can choose which to trust, with Bluesky choosing a default selection (maybe including theirs) for the majority who don't care
13,250 views
N
N
Replying to
@natanael
PGP web of trust has been known to be broken for years It's bad enough that I wrote a whole thread on PGP issues in 2019 (and it was old news by then) x.com/Natanael_L/s...
13,250 views
P
Replying to
@natanael
The Web of trust collapses as the diameter of the trust graph increases and the only way to avoid having the diameter become more than 4 is to have nodes of very high degree. Which means you need CA type parties involved but that doesn't mean you need to only use CAs.
13,250 views
Replies
P
Replying to
@phill
I wrote up my thoughts on this approach here. Basically the observation is that notarizing identity claims in a Merkle Tree type structure creates a work factor that quickly goes to infinity for backdating a claim. That is very useful. www.ietf.org/archive/id/d...
13,250 views