discussion: cw20-base methods should be nonpayable, or forward funds

[0xekez]

discussion: cw20-base methods should be nonpayable, or forward funds

[0xekez]

cc @webmaster128 / @ethanfrey . I'd be curious if this has been discussed before and your thoughts.

[hashedone]

That is a common problem with what to do when unnecessary tokens are sent to contract with the method, and I remember we discussed it with Ethan. Sure, you can add a check if there is anything sent with the method (which is equivalent to making it nonpayable, I even think there is a function named this way in cw-utils failing if anything is sent), but there is a still a thing - what does it prevent? If one wants for any reason to prevent sending tokens to contract, it is impossible - it can always be done via Bank::Send message unless specific chains have some own way of blocking incoming funds. Therefore it prevents nothing besides accidentally sending tokens between contracts. But then - how often are you manually sending a message to the contract and assigning tokens to it? In the vast majority of cases, you do it via some front-end app which has to know contract API anyway - therefore, there is just no need to add the possibility to assign native tokens in your front-end.

Adding the additional nonpayable call on every message seems unnecessary to me, makes additional code which I mostly don't care about (so decreases readability - not in a significant way, but still, not the direction I would like to take), it adds additional overhead (probably unnoticeably, but again - with no reason), with the only improvement is making it slightly easier to test manually.

While back in the day, I was convinced more to very strictly limit the ability to accept tokens by contract, right now, I would do it only when it actually introduces some problems - for eg., if I accept exactly one native token on instantiation and use it to determine "denom" my contract is operating on, then I would actually assert there is exactly one token send (as if they are many, I could pick the wrong one which may lead to problems).

[0xekez]

@hashedone, thank you for the thoughts. I used to also hold your point about code like this being un-needed as frontends are the API consumers. What I have come to appreciate over time is that as we are using non-restrictive licenses, and writing infrastructure-like code, there will be many frontends and many users who do unexpected things with this code.

I wonder if, when the result is funds locked in the contract, it is good to air on the side of caution?

[webmaster128]

I agree with @hashedone here. If we want such a solution, it should include bank sends. But x/bank does not know about the existence of x/wasm, which is also fair from a design point of view. One could imagine a hook system where x/wasm registers for such events. But this would create also some overhead. Something similar was discussed in CosmWasm/cosmwasm#1515 before.

[0xekez]

cool. thanks for the thoughts @hashedone and @webmaster128 :) sounds good to me.