4 ms·
> What would be the use case for "letter of the law" agreements Consider a service where you take money from a user, and then use that money to run a service o
by JoshTriplett 10y ago
> What would be the use case for "letter of the law" agreements
Consider a service where you take money from a user, and then use that money to run a service on their behalf, in a fully automated fashion. You get $x, you spend $y on a server and bandwidth, and you pocket x-y as profit. But if the user pays for your service on a credit card, they could dispute the $x, and leave you on the hook for $y, which could be arbitrarily large; you can immediately cancel their service, but that doesn't get you the $y you've already spent back. It'd be nice to have an enforceable agreement that "you can cancel at any time and stop paying us, at which point the service goes away, but under no circumstances will we owe you anything".
(This is also one reason why companies often don't offer discounts for paying up-front for a year or more, and credit card processors consider those high-risk: too many people dispute the transaction after a short time and get all their money back, including what they've already paid.)
I can imagine many more scenarios that rely on "once you have the money it can't go away", that don't work nearly as well with reversible transactions.
> how do you avoid massive problems due to buggy contracts as we're seeing?
In cases where you want human judgment as a safety precaution, write it in at the top level of the contract. "Notwithstanding anything else, all transactions are reversible for up to X days based on the judgment of Y, to be supplied via Z mechanism; if such judgment is not supplied, the transaction becomes immutable after those X days have passed." (Or, if you want, "all transactions can be disputed for up to X days, which will then hold them for judgment by Y for up to Z days".) The equivalent of "don't run untrusted software in ring 0"; the "pure letter of the law" space is ring 0, and many things could be built on one or more layers of indirection above that.
In cases where you don't want human judgment, then you avoid buggy contacts the same way you avoid buggy software: get a lot of people looking at it, and limit how much you trust things that haven't been heavily hammered on for years. Anyone investing millions into a "smart contract" today is taking an absurd risk.
You could also purchase insurance of some kind. If the price of that insurance is absurd, that should tell you something. But not every transaction should have to price in built-in insurance (and actuaries to properly price it).
- mikeash 10y agoI don't understand why current contracts don't suffice for that scenario. If the user disputes payment for services already rendered, you can sue them as things are now, and you'll win. It doesn't usually happen just because the cost is too high, but I don't see that this would change for "don't apply reasonableness, just read the letter of the agreement" contracts. Or just ask people to pay you in cash.
- JoshTriplett 10y ago> I don't understand why current contracts don't suffice for that scenario. If the user disputes payment for services already rendered, you can sue them as things are now, and you'll win. It doesn't usually happen just because the cost is too high, but I don't see that this would change for "don't apply reasonableness, just read the letter of the agreement" contracts. If you have to sue to recover the money, you're out the money until you win, and you might not win. You're also out the money if the entity you try to sue can't be sued, which can happen for a variety of reasons, such as "already pulled the money into a different jurisdiction you can't touch", or otherwise "doesn't have the money to recover". And in any case, it isn't worth the trouble unless the transaction is large enough, which is why most service businesses have to budget a certain amount for losses due to fraud. I'm not sure why you think an unbreakable agreement wouldn't fix that, assuming it was actually enforceable as such. > Or just ask people to pay you in cash. Cash doesn't function for online services (with very rare exceptions). And even for a cash transaction, people can sue you and demand their cash back. To clarify something: I don't think that the entire economy should use "rules as written" transactions; I think that would be an incredibly bad idea. For example, I want to store the majority of my savings somewhere that follows "spirit of the law", and has some protections and recourse for transactions I don't want. And I spend money on credit cards, confident that I can call up American Express if something goes horribly wrong. (I've never had to do so.) I do think it should be possible for me to intentionally and verifiably move a small amount of money from my bank account in a "rules as intended" system into a transactional account in a "rules as written" system, and then use it to conduct a transaction in that system. (Moving the funds is necessarily governed by "rules as intended" protections provided by my bank, as for example I would want some recourse from my bank if that transaction wasn't initiated by me. That then gives the bank an incentive to make absolutely sure they have valid authorization from me.)
- mikeash 10y agoI thought you were proposing a way of writing a normal paper (or PDF) contract which would have some legally-enforceable clause saying, "Don't try to interpret the intent of this, it means exactly what it says." To enforce such a contract in the event of a breach, you'd still get the usual legal machinery involved, but the court would skip the part where they try to apply reasonableness standards or figure out intent, and would just go straight to figuring out which party is correct and what the remedy should be according to the contract. Did I misunderstand the idea?