Most consensus protocols treat quorum as a cutoff—hit 67% and everything extra is noise. Dusk doesn't.
It allows multiple valid attestations for the same iteration, which means you can have different groups of validators all claiming agreement. The block certificate picks one. But that certificate isn't just a receipt—it names the exact validators who formed that attestation.
Why does that matter? Because rewards scale with votes beyond the quorum. The generator gets 70% of the block reward, plus a bonus up to 10% depending on how many extra attestations they pack in. So now there's a choice: finalize fast with the minimum, or wait a bit longer to scoop up more votes and capture that extra yield.
But here's the wrinkle: the certificate is constructed locally by each participant. There's no global canonical version. And once economic consequences depend on exactly who is in that voter set, the definition of that set becomes a consensus problem in itself.
So the protocol isn't just asking "did enough people agree?" It's asking "which specific people are we going to reward or penalize?" That adds accountability, sure. But it also adds a layer of complexity that could subtly shift incentives toward lower-latency validators who are more likely to be included.
I'm not convinced that tradeoff is clean. But I do think Dusk is onto something by refusing to treat extra consensus as irrelevant. Sometimes the interesting questions start where the protocol says "yes, but..."
@Dusk #dusk $DUSK
It allows multiple valid attestations for the same iteration, which means you can have different groups of validators all claiming agreement. The block certificate picks one. But that certificate isn't just a receipt—it names the exact validators who formed that attestation.
Why does that matter? Because rewards scale with votes beyond the quorum. The generator gets 70% of the block reward, plus a bonus up to 10% depending on how many extra attestations they pack in. So now there's a choice: finalize fast with the minimum, or wait a bit longer to scoop up more votes and capture that extra yield.
But here's the wrinkle: the certificate is constructed locally by each participant. There's no global canonical version. And once economic consequences depend on exactly who is in that voter set, the definition of that set becomes a consensus problem in itself.
So the protocol isn't just asking "did enough people agree?" It's asking "which specific people are we going to reward or penalize?" That adds accountability, sure. But it also adds a layer of complexity that could subtly shift incentives toward lower-latency validators who are more likely to be included.
I'm not convinced that tradeoff is clean. But I do think Dusk is onto something by refusing to treat extra consensus as irrelevant. Sometimes the interesting questions start where the protocol says "yes, but..."
@Dusk #dusk $DUSK
