Closed rootulp closed 5 months ago
FWIW I think this issue has existed since go-square was extracted from celestia-app. See https://github.com/celestiaorg/go-square/commit/68863fc876b175343fe68e21467c70652e629280#diff-f37b1db98350d69502a30f6f3dd88ae2f6844b2bb6e0c46f89d3b6a237788b3cR414
I just thought of a way to fix this bug without introducing an API breaking change, will publish an alternative PR soon.
Context
https://github.com/celestiaorg/go-square/blob/a84173978ed29049469dc5e84b4a58cfece98401/square/builder.go#L414
Problem
We want worstCaseShareIndexes to always use the upper bound for max square size (128) so that celestia-node doesn't have to be stateful when extending the ODS.
If
worstCaseShareIndexes
uses the max effective square size (64) then celestia-node would have to poll celestia-app for the gov max square size every block.Also note that celestia-app v1.x used a worstCaseShareIndexes of
128
so it was a consensus breaking change to drop down to64
.Proposal
Modify the second argument to this function to use the upper bound of
128