d@nny disc@hipsterelectron@circumstances.run
Sep 25, 2026, 9:02 PM@amy i'm more interested in data integrity at rest but therr are lots of really cool ways to do this
Replies
- d@nny disc@hipsterelectron@circumstances.runSep 25, 2026, 9:04 PM
@amy for an anonymous routing protocol i have the idea to do something like double ratchet/QUIC except with non-constant chunking, again to avoid replay attacks. this jives really well with again the same coding theory you describe above, where you can send not just raw chunks but erasure codes of such or suchlike
Sep 25, 2026, 9:11 PM@hipsterelectron lookup "set reconcilation with nearly optimal communication complexity" paper from the 90ies IIRC. It describes the algorithm designed to sync contacts for palm pilots and it's pretty neat.
Also I came up to RsTP when thinking about how TCP starts to fall apart in the face of really really huge latencies, like earth<>mars.- d@nny disc@hipsterelectron@circumstances.runSep 25, 2026, 9:14 PM
@amy TCP is vint cerf's replay-attack-as-a-service and is remarkably unserious. QUIC is quite literally just the signal protocol (replete with trevor perrin as consultant) but being the internet they of course have to fuck it up for the inane "header compression" bullshit
- d@nny disc@hipsterelectron@circumstances.runSep 25, 2026, 9:16 PM
@amy that paper has the jane street guy as first author how curious
- d@nny disc@hipsterelectron@circumstances.runSep 25, 2026, 9:17 PM
@amy ughhhhh i'm so over polynomials
- Sep 25, 2026, 9:20 PM
@hipsterelectron I mean, I did say that it came from corpo land. Polynomials they have their flaws, yes, but also it's interesting work for it's time and arguably I think there's unexplored possibilities with the shape of the technique they describe
- d@nny disc@hipsterelectron@circumstances.runSep 25, 2026, 9:44 PM
@amy i just think they're overused i don't really have any problems with them. this was just whining
- Sep 25, 2026, 9:24 PM
- d@nny disc@hipsterelectron@circumstances.runSep 25, 2026, 9:43 PM
@amy i am uninterested in factoring weaknesses because everyone should be using pure diffie-hellman instead
- d@nny disc@hipsterelectron@circumstances.runSep 25, 2026, 9:55 PM
@amy working on this tonight in fact
- Sep 25, 2026, 9:17 PM
@hipsterelectron tell me more about this issue with TCP? I can see how it can be vulnerable to a replay exactly one before RST is sent because of a state machine violation.
- d@nny disc@hipsterelectron@circumstances.runSep 25, 2026, 9:42 PM
@amy any intermediate can choose to drop a packet and force the sending party to retransmit. now that i think about it though you're right that's not a replay attack a replay attack would be from the reuse of a TLS session
Emily_Semily_s@mastodon.me.ukSep 25, 2026, 9:14 PM@amy @hipsterelectron you'll not like the sound of the tiptop (Taking IP To Other Planets) working group then :P
- d@nny disc@hipsterelectron@circumstances.runSep 25, 2026, 9:19 PM
- d@nny disc@hipsterelectron@circumstances.runSep 25, 2026, 9:20 PM
- d@nny disc@hipsterelectron@circumstances.runSep 25, 2026, 9:22 PM
- Sep 25, 2026, 9:25 PM
@hipsterelectron @emily_s links are totally distinct from addressing. A link is literally just the establishment of shared state.
- d@nny disc@hipsterelectron@circumstances.runSep 25, 2026, 9:36 PM