avi pfp
avi
@avichalp
i think, it is enough if the light nodes can sample the state and verify the execution proofs
1 reply
0 recast
1 reaction

androidsixteen pfp
androidsixteen
@androidsixteen.eth
IIUC, I think Vitalik discussed this in this piece under the "Light clients and fraud proofs" section: https://vitalik.eth.limo/general/2020/08/17/philosophy.html In this scenario, who "raises the alarm"? I think we still need entities running fully verifying nodes to monitor the network and flag issues
1 reply
0 recast
2 reactions

avi pfp
avi
@avichalp
he also mentions this scenario that i described above. as long as data is published, and sampling nodes collectively ensure that it is available then the validity proofs of execution are good enough
1 reply
0 recast
1 reaction

androidsixteen pfp
androidsixteen
@androidsixteen.eth
Is this the Celestia approach via DAS? Also doesn’t this require sharding, as you need a DA layer and various execution layers each with their own honest minority? That seems to be a different posture from Solana
1 reply
0 recast
0 reaction

avi pfp
avi
@avichalp
i think, in celesta you would still need full nodes to generate fraud proofs but they are working on removing that constraint with Zoda https://forum.celestia.org/t/using-zoda-in-celestia/1863 yup! i don’t think solana is trying to do this. i was just thinking of a hypothetical L1 with these two properties
1 reply
0 recast
1 reaction