Proposal: Remove the PartSetHeader from the Header and don't Sign over it when Voting

[evan-forbes]

Summary

Since we use the DataAvailabilityHeader to commit to block data, the PartSetHeader is redundant and should be removed from the header so it is not included in the header hash or signed over.

Details

As described in ADR 007 #504, we are treating our fork of tendermint like a black box and minimizing changes. This means reverting the changes we've made so far to get rid of the PartSetHeader (see #509 and #510). Despite this, we still don't want to sign over the PartSetHeader or include it in the header hash, as it is a redundant commit to the block data. Previously, this was already done in #456, which removed the PSH from the header and the canonical vote. We can no longer use that mechanism to remove the PSH from the header, since we still need the BlockID and are reverting #441 in #510.

Proposal

Replace the BlockID struct with a simple hash only in header.LastBlockID and CanonicalVote.CanonicalBlockID. Notably, we will also likely have to change some BlockID methods to not use the PartSetHeader such as Equals, ValidateBasic, and Key.

Implementation

• replace header.LastBlockID with a simple hash
• replace CanonicalVote.CanonicalBlockID with a simple hash
• investigate whether or not we have to alter certain methods of BlockID to use the PSH

References

ADR #504
Blocked by #510
Blocked by #492

[evan-forbes]

this is no longer planned afaiu, please reopen @liamsi or @adlerjohn if this is not the case