John Adler

Also related: https://github.com/lazyledger/lazyledger-specs/issues/110#issuecomment-818786351

One downside of using shares is that, depending on the number of bytes used to encode the length delimiter as varint, the number of shares could change. While the number...

I'm not convinced yet of an attack vector that requires length delimiter per share instead of per data.

How would you determine if a share is the first share of a namespace, under that proposal?

This is now unblocked @iofq!

> 8 bytes additional Namespace for each share via NMT wrapper > Can be avoided if we enable once Why can't it be avoided today? Storage format is an implementation...

Thanks for working through this @iofq! Good work!

Are you proposing a consensus-critical change (i.e. having multiple data commitments per PFD), or an application-level standard?

Such an overhead doesn't necessarily have to be enforced by the protocol, but can rather be an emergent property of a separate fee market for Celestia-app transactions: https://github.com/celestiaorg/celestia-app/issues/738

To clarify: 1. the share size that is erasure coded is still `SHARE_SIZE` 2. `SHARE_SIZE - NAMESPACE_ID_BYTES - SHARE_RESERVED_BYTES` is how requests with a reserved namespace ID are chunked (txs,...