DSys RG June 2022: Making CRDTs Byzantine Fault Tolerant

Paper: Making CRDTs Byzantine Fault Tolerant

00:07:37 Ali: Thanks, let me increase the microphone gain.
00:14:42 Brendan O’Brien: Fantastic summary!
00:14:51 Philipp Krüger: 100% ^
00:21:38 Philipp Krüger: It’s on the bottom left paragraph on the shown page
00:24:39 Brendan O’Brien: “trust but verify” :slightly_smiling_face:
00:24:57 Brian: Sounds like a possible latency win
00:25:03 Brendan O’Brien: sorta. byzentine nodes can begin faulting at any point
00:33:01 Brendan O’Brien: or maybe all messages you have accepted
00:33:39 Brian: I have to leave early. Pleasure to chat with you all today!
00:36:11 Brendan O’Brien: ahhhhhhhhh o
00:36:29 Brendan O’Brien: *ok, so each peer retains a bloom filter for their other peers?
00:36:58 Marc-Antoine Parent: there’s such a thing as a growable bloom filter, right?
00:37:13 Philipp Krüger: I mean, a gig more when there’s a maybe some TB of messages? :stuck_out_tongue:
00:39:09 Brendan O’Brien: scalable bloom filters: https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&ved=2ahUKEwiEvo_79b74AhVGFlkFHUG2CQIQFnoECAQQAQ&url=https%3A%2F%2Fhaslab.uminho.pt%2Fcbm%2Ffiles%2Fdbloom.pdf&usg=AOvVaw1b-xw_uhg6iqyN6L7b2tWU
00:40:16 Na: i’ll be a minute
00:40:24 Na: as i debug audio issues >_<
00:40:29 Philipp Krüger: >.<
00:41:57 Na: > Byzantine nodes may attempt to cause a performance degradation by generating a large number of concurrent updates, and hence a large number of heads, or updates with a large number of predecessor hashes, but these updates will not affect the correctness of the algorithm.
00:42:29 Brendan O’Brien: have to jump unfortunately. Fantastic chat!
00:42:38 Na: \o
00:49:13 Juho Myllylahti: Thank you for the paper recap and the interesting discussion. I’ll need to leave unfortunately, but see you next month :slight_smile:
00:49:20 Marc-Antoine Parent: [1805.04263] OpSets: Sequential Specifications for Replicated Datatypes (Extended Version)
00:49:30 Philipp Krüger: Thanks
00:50:25 Marc-Antoine Parent: (I would very much like that paper to be one of the next)
00:51:52 Marc-Antoine Parent: That is a nice benefit vs vector clock
00:59:43 Na (Sodium): it’s like a torrent where there’s a missing chunk in the middle
00:59:51 Brooklyn Zelenka (@expede): Exactly
01:05:18 Ben Kircher: Thanks everyone!
01:05:24 Brooklyn Zelenka (@expede): Thanks Ben!
01:06:38 Owen Bickford: Gotta jump. This was a super interesting convo. :wave:t2:
01:06:50 Na (Sodium): \o
01:07:16 Ali: I’ve got to go, thanks a lot.