Remove the PartSetHeader where possible

[evan-forbes]

Summary

Remove the PartSetHeader.

Details

As mentioned in #184, we now are propagating block data via IPFS and no longer need to do so via the PartSetHeader. After #218 and #374 are completed, we will have the necessary groundwork finished to actually remove the PartSetHeader completely. Complete removal is the next step to moving on with replacing the BlockID with the Header hash.

Action Items

• Decouple PartSetHeader from BlockID #441
• Don't sign over the PartSetHeader or include it in the Header #456
• Remove from other structs/figure out exactly where it is needed

References

part of completing #184

[Wondertan]

#293 is also highly related. PartSetHeader is used extensively in consensus reactor and it might be necessary to implement block propagation first before removing PartSetHeader

[liamsi]

and it might be necessary to implement block propagation first before removing PartSetHeader

It might also be possible to simply use the existing tendermint block propagation mechanism (maybe using erasure coded shares instead of current chunking) to unblock this and use the DAHeader instead of partsetheader where needed.

[liamsi]

ref: #389 #423 #184

[evan-forbes]

Update: Instead of removing the PartSetHeader completely, will still use it, or something similar (#427 & #434), to propagate blocks during consensus. Regardless, we should still follow through with removal of the PartSetHeader where possible. Starting with #456.

[evan-forbes]

Update: Removing the PartSetHeader from Vote and Commit would require a refactor in the consensus reactor to change how the PSH is passed. It should be noted that while the PartSetHeader is still included in Vote, it is no longer signed over since #457

Removing the PSH from Commit is doable without a large refactor, see https://github.com/celestiaorg/lazyledger-core/tree/evan/remove-psh-4, but would require significant changes to tests. I think we should continue with the removal of the PSH from the Commit after we update to a newer version of tendermint.

[liamsi]

I think we should continue with the removal of the PSH from the Commit after we update to a newer version of tendermint.

Yeah, I agree there is no rush and it makes sense to update to a newer tendermint version first.

[evan-forbes]

we have since decided to keep tendermint as a black box, and complete removal of the PSH is no longer an option. That being said, we still have plans to experiment with removing it from the header and not signing over it #511