robert

> Are Views also tracking nodes assigned roles? We'd presumably send initial backing statements only to other backers, and only gossip double backing statements to everybody, for example. They don't...

Out of curiosity, what is the heaviest part of GRANDPA? Can we optimize GRANDPA lower-level to achieve similar results without increasing the finality time?

This might all be irrelevant if we just have backing groups of size 1 or 2. Groups of 3 or more are likely to benefit from this.

Ladi and I also had a longer discussion about the protocol and decided that we don't need anything other than regular merklization as I described in the original post. >...

For us to make use of the systemic code effectively we'd have to change the validator indices who hold the systemic chunks more often than we do now, which is...

The first two `validator_index + parachain_id` we can do easily in approval-checking, but the last part `relay_parent_slot_no` isn't as easily available especially for something like remote disputes. We could have...

Please add some documentation with examples to the crate

I think we'd want 50% master, 50% prev-release with a few parachains registered. And then maybe 50% of collators on master, 50% on prev-release. And then we'd have some health...

The companion kind of needs to go in first, before this PR ever lands and could be released in a Polkadot/Kusama runtime. We'd also need all parachains to upgrade their...

Yeah, we should try a branch with those commits removed. 5b74841ea76195fb804a02c0336d05d9a2d3091b , 3ae71a552256273c6a82c3824451fb73423b2232, 7311a3aa9f83730a079b0899c34bb813dd96c4e1 all seem like viable culprits.