Conversation

UPDATE TO THE UPDATE: We're shipping a bloom filter after all! Thanks for giving me the opportunity to debug Bloom filter false positives,

@martinkl

Quote Tweet

Peter van Hardenberg

@pvh

· Sep 25, 2019

Update: Nothing came of this. twitter.com/pvh/status/979…

Show this thread

Replying to

and

Still roughly as described in martin.kleppmann.com/2020/12/02/blo? I may be using this same also!

10:51 PM · Apr 20, 2021Twitter Web App

Replying to

and

I just read this link. Wow, I’m currently using merkle trees with hashes for this problem, getting a window of messages to send but it easily over sends. I may need to try this out too!

James Long

@jlongster

Apr 21, 2021

I wonder how large the bloom filter gets though with tens of thousands of messages. I prune the merkle tree, losing accuracy but it stays very small as messages scale

Show replies

Replying to

and

Quite so -- we have a pretty decent two-node hash graph sync working now. I expect it to extend really nicely to n-peer & incremental sync though we have more work to do for that.