Conversation

Notices

1. mia (mia@freespeechextremist.com)'s status on Sunday, 29-Oct-2023 04:55:23 JST mia

   in reply to

   • jaf
   • Izuru Yakumo (八雲 出流)
   @jeff @yakumo_izuru
   
   And what would be the already finished component in place of that struct snippet he shared?
   • jaf (jeff@misinformation.wikileaks2.org)'s status on Sunday, 29-Oct-2023 04:55:23 JST jaf

     in reply to

     • Izuru Yakumo (八雲 出流)
     @mia @yakumo_izuru to solve the problem of activity pub censorship? a gossipy LRU cache for activitypub objects and the resources they reference which utilizes a decaying bloom filter to lessy the talky. i feel like i could make a toy in elixir but i dont want to make this a "Win internet argument" project. i'd want it to be actually useful and solve that problem. i have been lurking DHT research and key space flappy networks are a bad time, this will be made flappy af as it would be trivial to shitflood. it makes a lot more sense to not use a DHT but do something closer to what ygg 0.5 does.
     
     ygg does a route cache with gossip using bloom filters to dampen amplified propigation. it really looks like it'd scale to the size fedi would need as instead of gossiping a route you are gossiping arbitrary data, be it an apub object or some relatively small small file it links to. a DHT will not scale to a size of millions like this needs to, but a distributed LRU cache would.
     
     everyone has opinions on why we should DO X as do I, i just want to present some data that suggests the popular meme isn't good enough at scale and there is research that shows a better way. then we have p who wants to muck around with data formats not solving the high level problem, then getting mad when people point it out.
      likes this.
   • mia (mia@freespeechextremist.com)'s status on Sunday, 29-Oct-2023 04:55:25 JST mia

     in reply to

     • jaf
     • Izuru Yakumo (八雲 出流)
     @jeff @yakumo_izuru
     
     What tool do you specifically think he is needlessly re-making?
   • jaf (jeff@misinformation.wikileaks2.org)'s status on Sunday, 29-Oct-2023 04:55:25 JST jaf

     in reply to

     • Izuru Yakumo (八雲 出流)
     @mia @yakumo_izuru revolver
   • jaf (jeff@misinformation.wikileaks2.org)'s status on Sunday, 29-Oct-2023 04:55:26 JST jaf

     in reply to

     • jaf
     • Izuru Yakumo (八雲 出流)
     @yakumo_izuru mainly irked that p still has yet to learn this lession.
   • jaf (jeff@misinformation.wikileaks2.org)'s status on Sunday, 29-Oct-2023 04:55:27 JST jaf

     in reply to

     • Izuru Yakumo (八雲 出流)
     @yakumo_izuru that is the solution now yes. it was not an option in 2017
   • Izuru Yakumo (八雲 出流) (yakumo_izuru@misskey.m544.net)'s status on Sunday, 29-Oct-2023 04:55:28 JST Izuru Yakumo (八雲 出流)

     in reply to

     • jaf

     @jeff@misinformation.wikileaks2.org lol or use QUIC ;P

   • jaf (jeff@misinformation.wikileaks2.org)'s status on Sunday, 29-Oct-2023 04:55:29 JST jaf

     in reply to

     • jaf
     if you need to create a new binary protocol make it serialized as ASN.1/DER and be done with it.
     only make a new binary protocol if you cannot fit your requirements into any existing protocols.
   • jaf (jeff@misinformation.wikileaks2.org)'s status on Sunday, 29-Oct-2023 04:55:30 JST jaf
     lokinet had a nih binary udp wire protocol and fucking with it sapped 2 years from development. should've used dtls like my gut said to.