@dcz First time hearing about #Kademlia and #DHT. It's fascinating that #IPFS was created using DHT.

@snowmead Rewriting a #Kademlia #DHT implementation. Despite there being like 10 different crates, none are at the same time low-level, with builtin network module, and not tied to a specific network library. Not to even mention testable or other luxuries like being generic.

Really cool video by @b5 looking at optimizations over #kademlia #dht algos by taking latency into account when searching for data. Seems like it would be very useful for #p2p and #localfirst protocols.

https://www.youtube.com/watch?v=-OUUXIldZas

@salamander Wobei mit etwas #Kademlia müßte das auch verteilt gehen.

Bei mir ist heut ist echt #nerdsniping angesagt.

Kademlia: https://de.wikipedia.org/wiki/Kademlia

I wonder what is the impact of caching on mutable values in #Kademlia-like #DHT.
I mean, the caching algorithm is perfectly sound for constant values, but mutable values are not mutated on all peers having stored that value. So, if my understanding is correct, this results in view fragmentation.
I suppose that the re-publish algorithm could solve this with a smart merge strategy depending on the nature of the published data, but that sounds fragile.
Insights?