Most Chains Shout. Dusk Uses a Map.
networking is the layer nobody writes about and it decides more then people think.
Kadcast is Dusks P2P layer, and the docs describe it as a structured overlay instead of random gossip, chosen to reduce bandwidth and improve latency predictability.
gossip is how most networks propagate. a node hears something, tells its peers, they tell theirs. it works and its robust, but its wasteful, the same message arrives at the same node repeatedly, and arrival time varies.
a structured overlay routes by position instead. nodes know where things should go rather then flooding outward and hoping.
heres the word doing the work though, and its not speed. its predictability.
a market venue cares less about the fastest possible case then about the spread between best and worst. a system that averages fast but occasionaly takes much longer is harder to build settlement guarantees on then one thats consistently adequate.
so Kadcast being chosen for latency predictability lines up with everything else here, deterministic finality, defined settlement stages, rules that execute rather then get reviewed. the whole stack keeps picking certainty over peak performance.
is optimising for predictability instead of speed the right call for market infrastructure, or does it cost you the ceiling you'd need later?
Predictable, or capped?
#dusk @Dusk $DUSK
networking is the layer nobody writes about and it decides more then people think.
Kadcast is Dusks P2P layer, and the docs describe it as a structured overlay instead of random gossip, chosen to reduce bandwidth and improve latency predictability.
gossip is how most networks propagate. a node hears something, tells its peers, they tell theirs. it works and its robust, but its wasteful, the same message arrives at the same node repeatedly, and arrival time varies.
a structured overlay routes by position instead. nodes know where things should go rather then flooding outward and hoping.
heres the word doing the work though, and its not speed. its predictability.
a market venue cares less about the fastest possible case then about the spread between best and worst. a system that averages fast but occasionaly takes much longer is harder to build settlement guarantees on then one thats consistently adequate.
so Kadcast being chosen for latency predictability lines up with everything else here, deterministic finality, defined settlement stages, rules that execute rather then get reviewed. the whole stack keeps picking certainty over peak performance.
is optimising for predictability instead of speed the right call for market infrastructure, or does it cost you the ceiling you'd need later?
Predictable, or capped?
#dusk @Dusk $DUSK