PBTS: evidence

[josef-widder]

With proposer-based time, the consensus messages will change. In particular, there will be possible misbehavior w.r.t. the time value that is part of the messages. We need to capture in the specification the new evidence validation rules.

This might also be a good opportunity to add "double proposal evidence", that is, if the proposer equivocates.

[cason]

Will the misbehavior here be proposing values with time in the future? How can we "prove" that?

[cmwaters]

Was there any decision made on this? Shall we keep the issue around for future reference?

[josef-widder]

@cason, I think the idea was to punish if the time in the propose message and in the block (sent by a faulty proposer) don't match. Right?

I guess it makes sense to keep the issue open for now. Also we are currently working on accountability for amnesia, so it might make sense to think more broadly about different evidence types.

[cason]

We can prove two "atemporal" misbehaviors for PBTS:

1. Proposing a value (BlockID) with a timestamp that does not match the block's timestamp (mentioned above);
2. Proposing at height h a timestamp smaller than or equals to the timestamp of the block committed at height h-1, thus violating monotonicity.

To prove other misbehaviors, we would need a way to produce "certificated" timestamps, which is more complex.