Cosmos-sdk: Consider whether or not we want to trust the full node by default for post commands

[cwgoes]

e.g. sending a transaction

I think perhaps not - otherwise a full node could lie about sequence number, account number, etc.

cc @HaoyangLiu

ref https://github.com/cosmos/cosmos-sdk/pull/2210/files#diff-38b4a4a9824366646939d96f3dc02184L65

[ValarDragon]

I thought the plan was to do merkle proofs for account querying? You would then have sequence number, account number, etc.

[rigelrozanski]

Well - the sender needs to sign the message anyways - so if the signed message doesn't make the sequence number amounts etc. the message would just fail (would fees be deducted however? - I don't think so because we check the signatures before deducting the fees)

meh, so I don't see how this is really an issue, the worst a node could do would be not broadcast your transaction

[ValarDragon]

Fees won't be deducted if the sequence number is invalid, so I agree, this does seem like a non-issue as long as your connected to multiple full nodes.

[jackzampolin]

Closing this as the lite client proof system is undergoing a refactor.

[cwgoes]

Closing this as the lite client proof system is undergoing a refactor.

We should explicitly consider this griefing vector though.

[alexanderbez]

What is the griefing vector? Spam?

[cwgoes]

What is the griefing vector? Spam?

Getting the client to sign an invalid transaction by returning the wrong current account nonce or account number.

[jackzampolin]

Yes, but I don't think this issue is the right spot to track that. Going to close.

[rigelrozanski]

Where's the best place to track that?

[jackzampolin]

Currently writes default to trust the node and queries default to not trust. Going to go ahead and close this issue. If reopening, please add an actionable item.