Make antehandler local fee check more robust

[ValarDragon]

There ought to be a value metric per validator, which says this coin denomination is worth x gas to me. Then we should change ensureSufficientMempoolFees to instead calculate the gas worth of the provided fees, and compare it to the gas wanted.

This is tagged pre1.0 since we only have one coin at launch. This only becomes a problem once there are multiple coins.

[alexanderbez]

See #3101 (comment)

[jackzampolin]

We just fixed an issue here right? Or are we waiting on #3248 to close this?

[alexanderbez]

I believe @ValarDragon needs to further elaborate on the exact functions he's proposing. #3248 lays the groundwork though for sure.

[ValarDragon]

The user provides how many gas 1 coin of each given denom is worth. Then you do the linear combination to determine the amount of gas for which you think this set of coins has paid for. (paid_gas = \sum denom_amount * denom_gas_worth) You then check that the paid for gas is >= to gas wanted.

This can easily be extended to allow the validator to charge for gas wanted in their local checkTx calculation with the same idea.

[rigelrozanski]

@sunnya97 weren't we talking about some form of decentralized value prioritization per different coin type? aka each validator would frequently post updates (via a tx) on what their valuation of each token is relative to the atom, and then the median value would be taken as the global valuation to be used for gas prioritization? - Is this ringing any 🔔 s guys?

[sunnya97]

@rigelrozanski Yes, that is what we want long term, but we decided that for launch, we're just doing locally set mimimum GasCost at the node level (not in consensus).

@ValarDragon If I understand this correctly, this is regarding being allowed to pay fees using multiple denoms in the same tx? For simplicity, I think we can allow Fees to be paid in only one denom for now.

[ValarDragon]

Since were having multiple coins on testnets already, shouldn't we go ahead and fix it given the confusion in GoS? (I think it will be easier to fix this, than it will be to explain the currently really weird behavior to our validators and validators of other chains)

I'm not that opinionated on this, so feel free to change it back to pre1.0 tho