Mempool should include replay protection

[rigelrozanski]

Aka - check and increase the sequence number within the mempool transactions. If a block has not yet been submitted, a transaction could be rejected by the mempool if there is an unconfirmed transaction in the mempool with the same signers and sequence number

[ethanfrey]

Sure. Seems good. I think ethermint had lots of trouble with nonces, and out of order or something...

Don't know what they did, but if this is the solution for it, we should do it, so rigel:5 and rigel:6 could both pass checktx in the same block

@ebuchman is this how to deal with the ethermint nonce issue?

[ebuchman]

What am I missing? Mempool calls CheckTx which should handle replay protection.

Mempool also keeps a cache and rejects duplicate transactions that its seen. Cache is only like size 10K so eventually it rolls over and allows old duplicates ... thats where CheckTx protection really shines

[rigelrozanski]

@ethanfrey @ebuchman I think we're on the same page based on these comments - This issue was regards to some design decisions being made in the unstable, which is consistent now with maintaining replay protection within checkTx, I believe was already like this in 0.6.
@ebuchman were you being sarcastic about CheckTx shining about replay protection being faulty for old old replays? this sounds like an new issue, if it's relevant to basecoing please open a new issue

[ebuchman]

no sarcasm. just saying tendermint mempool tries to do basic replay protection, but doesnt guarantee it. need CheckTx for that