ICS2: Update for changes to client state verification

[cwgoes]

Thanks @AdityaSripal - cosmos/cosmos-sdk#6524 (comment).

[mpoke]

@cwgoes Is this issue still relevant?

[colin-axner]

I think so, I don't see the check we do in ibc-go represented in the spec. ValidateSelfClient is verifying that the client state stored on the counterparty chain has the correct chain specific fields. See our func

[mpoke]

I think so, I don't see the check we do in ibc-go represented in the spec. ValidateSelfClient is verifying that the client state stored on the counterparty chain has the correct chain specific fields. See our func

Thanks for the reply. So every client type should define ValidateSelfClient, right?

[colin-axner]

No each chain must define ValidateSelfClient. It probably makes more sense for this function to exist at 03-connection. Does that make sense?

[mpoke]

No each chain must define ValidateSelfClient. It probably makes more sense for this function to exist at 03-connection. Does that make sense?

ValidateSelfClient would contain specific logic for every client type. For example, the ibc-go function contains Tendermint logic. I agree that the function is called from ICS 3, but it should be defined in ICS 2 in an abstract way, and the specific logic should be added to every client spec.

[colin-axner]

I suppose that makes sense from a specification point of view. In our codebase it doesn't make sense (ibc-go doesn't need a solo machine ValidateSelfClient function)

[colin-axner]

Can this issue be closed now? ref #802

[mpoke]

It seems so.