Egregoros

@i @p @lain @silverpill @jaff @cjd @mint @sun >fedi is simply not interesting enough for spam and abuse
it doesn't matter if it's interesting enough with more time that goes on more LLMs will come online, users will have chips in their phones capable of running them. eventually fedi won't even be able to work as intended because one bad actor decided to send an agent swarm to attack every instance's security, co-mingle with fedizens to prevent discourse, and autonomously OSINT users to dox them

Yeah, this is probably the weakest part of the SLUFI spec.

> The announce endpoint must be rate-limited with regard to the announcement of never before seen nodes so as to limit resource exhaustion denial of service through the creation of fake nodes. A node should not be considered “never before seen” unless its last DSR has expired more than 1 week ago. Every node must also impose a maximum DSR count, and when this number is reached, no DSRs of nodes not currently in the database are accepted at all, either by announcement or gossip. Finally, nodes accepting announcements must offer some method of blacklisting such that invalid nodes can be permanently excluded. Additional limits such as maximum nodes per second level domain are also recommended.

There isn't much else you can do here, but someone who wants to be a dickhead can just sit there and keep the announcements at rate limit.

MAYBE, I could put a hashcash option on the /announce endpoint. I'm thinking I'll probably write the trusted/gossipper server as a standalone, and then the pleroma patch will just be as a leaf node

@r000t @cjd @p @lain @nobody @silverpill @jaff @mint @sun
> spam, sitters
unsolved/unsolvable. reputation nets work but then have the failure mode that actually-new people don't have them. cost hasn't worked super great (people sit domains like with the iphone) for icann.

proof of work for zero-rep links and then not enforcing it for trusted links, maybe, pushes the checking burdens down to people that actually want to run bastions (gmail used to get fiat bonuses in spam rules because gmail ran outbound spam checkers in the early days)

Fortunately we're really only worried about resource exhaustion, and this is considering single digit KB records, furthermore administrators can purge attack domains from their db and that will propagate to anyone who discovered those records from them.

I would actually wanted to switch it from gossip to lookup (like DNS) but it currently allows for trust loops and recursive lookups through a loopy trust network would become really complicated... Still, this is a weak area of the spec so it might be worth re-visiting...

Maybe to gossip bloom filters and then only recurse a lookup on match...

SLUFI is published!

https://thegoodwork.substack.com/p/slufi

A Simple, Lightweight, Unopinionated, Federation Interconnect.

The problem: While numerous network transports exist (cjdns, Yggdrasil, I2P, Tor, etc), and numerous name services exist (Namecoin, ENS, Unstoppable Domains, PKT, etc), almost all federation takes place over ICANN+IP because we currently have no way to federate across network boundaries, and ICANN+IP is the only thing that everybody has.

I claim that there is no possible solution SIMPLER than SLUFI, except with central authority or PRECIPITOUS loss of security. Furthermore, I claim there is no possible solution that is more secure than SLUFI, except with something like a blockchain and a PRECIPITOUS loss of simplicity.

So while I do not think this protocol is complete (yet), I do think SLUFI is the only realistic starting point for reaching a solution to this problem.

Tagging some people who might be interested:
@sun
@lain
@p
@mint
@jaff
@silverpill

I probably forgot a bunch of very important people, apologies in advance...