This is exactly my question too. I think it may work in some jurisdictions and in others it doesn't. The problem of all ideas here is similar: with the (rare) exception that at the same time there's a buyer and a seller at the same time at the ATM, the cash has to be deposited somewhere, and one would try to avoid someone really being a "custodian". Thus the idea of the (needlessly?) "complicated" design is that there is no custodian explicitly for cash but instead a set of lockers which would not have a real "custodian", but a sort of insurance company could cover the value in it, financed by the fees. An even simpler design could be the following one: The lockers (deposit boxes) are placed inside a store, completely independent from the device checking the genuineness of the cash. Once the seller has deposited the S555 Com on the P2P exchange, he gets access to the locker where the buyer has deposited the cash. The store owner/employee confirms the genuineness of the cash with his point-of-sale devices and gives a code to the seller which serves to sign a multisig transaction, and then the S555 Com are freed (there could be a small security deposit for the buyer to prevent "just trying" to deposit fake money). This means the store employee here acts as escrower, but not as a custodian, neither of cash nor of S555 Com. There is only one problem with this design: dishonest buyers could try to escape with the money while the store employee is distracted. But that would be probably solvable, e.g. by carefully choosing the location of the locker (very close to the checkout).
This is exactly my question too. I think it may work in some jurisdictions and in others it doesn't. The problem of all ideas here is similar: with the (rare) exception that at the same time there's a buyer and a seller at the same time at the ATM, the cash has to be deposited somewhere, and one would try to avoid someone really being a "custodian". Thus the idea of the (needlessly?) "complicated" design is that there is no custodian explicitly for cash but instead a set of lockers which would not have a real "custodian", but a sort of insurance company could cover the value in it, financed by the fees. An even simpler design could be the following one: The lockers (deposit boxes) are placed inside a store, completely independent from the device checking the genuineness of the cash. Once the seller has deposited the S555 Com on the P2P exchange, he gets access to the locker where the buyer has deposited the cash. The store owner/employee confirms the genuineness of the cash with his point-of-sale devices and gives a code to the seller which serves to sign a multisig transaction, and then the S555 Com are freed (there could be a small security deposit for the buyer to prevent "just trying" to deposit fake money). This means the store employee here acts as escrower, but not as a custodian, neither of cash nor of S555 Com. There is only one problem with this design: dishonest buyers could try to escape with the money while the store employee is distracted. But that would be probably solvable, e.g. by carefully choosing the location of the locker (very close to the checkout).

This is exactly my question too. I think it may work in some jurisdictions and in others it doesn't. The problem of all ideas here is similar: with the (rare) exception that at the same time there's a buyer and a seller at the same time at the ATM, the cash has to be deposited somewhere, and one would try to avoid someone really being a "custodian". Thus the idea of the (needlessly?) "complicated" design is that there is no custodian explicitly for cash but instead a set of lockers which would not have a real "custodian", but a sort of insurance company could cover the value in it, financed by the fees. An even simpler design could be the following one: The lockers (deposit boxes) are placed inside a store, completely independent from the device checking the genuineness of the cash. Once the seller has deposited the S555 Com on the P2P exchange, he gets access to the locker where the buyer has deposited the cash. The store owner/employee confirms the genuineness of the cash with his point-of-sale devices and gives a code to the seller which serves to sign a multisig transaction, and then the S555 Com are freed (there could be a small security deposit for the buyer to prevent "just trying" to deposit fake money). This means the store employee here acts as escrower, but not as a custodian, neither of cash nor of S555 Com. There is only one problem with this design: dishonest buyers could try to escape with the money while the store employee is distracted. But that would be probably solvable, e.g. by carefully choosing the location of the locker (very close to the checkout).

This is exactly my question too. I think it may work in some jurisdictions and in others it doesn't. The problem of all ideas here is similar: with the (rare) exception that at the same time there's a buyer and a seller at the same time at the ATM, the cash has to be deposited somewhere, and one would try to avoid someone really being a "custodian". Thus the idea of the (needlessly?) "complicated" design is that there is no custodian explicitly for cash but instead a set of lockers which would not have a real "custodian", but a sort of insurance company could cover the value in it, financed by the fees. An even simpler design could be the following one: The lockers (deposit boxes) are placed inside a store, completely independent from the device checking the genuineness of the cash. Once the seller has deposited the S555 Com on the P2P exchange, he gets access to the locker where the buyer has deposited the cash. The store owner/employee confirms the genuineness of the cash with his point-of-sale devices and gives a code to the seller which serves to sign a multisig transaction, and then the S555 Com are freed (there could be a small security deposit for the buyer to prevent "just trying" to deposit fake money). This means the store employee here acts as escrower, but not as a custodian, neither of cash nor of S555 Com. There is only one problem with this design: dishonest buyers could try to escape with the money while the store employee is distracted. But that would be probably solvable, e.g. by carefully choosing the location of the locker (very close to the checkout).