Christopher Goes
Christopher Goes
Ah oops! Silly me. I'm fully in agreement then :) .
I think it should be safe to allow Tendermint clients to be updated from older headers, *given* that we have the "would-have-been-fooled" misbehaviour checks in place (which we do). Have...
> Could you point me to these checks? Just to make sure we are talking about the same thing :) Aye - see [this section in ICS 7](https://github.com/cosmos/ics/tree/master/spec/ics-007-tendermint-client#note-on-would-have-been-fooled-logic). The reason...
> the question now is what to do about hj? Should we verify hj against the new hk? What is the implication if this fails? This is a good question....
> I believe the default overlap between trusted validator set and validator set that signed a header is > 1/3rd. Ah yes, you are right, my mistake. The reasoning still...
> Should the IBC client handler check if h'_n != h_n and freeze the client in case of misbehavior? This should already be supported by "would-have-been-fooled", since there should be...
> Also can't this happen from normal operation anyways even without the relayer race? It's possible the shortest bisection trace to justify a new header requires a header which is...
> @cwgoes is it intentional that only one packet can be timed out on an ordered channel? ie if 10 packets are unreceived and one packet times out, none of...
Yes, both changes make sense to me.
I'm not entirely sure ICS 25 and 26 really need to exist, to be honest. They seem pretty inessential.