At question is a promise that purports to recursively create and cash a promise
with identical text, causing G's vote on proposal 8543 to fluctuate
indefinitely between two values.
# What does it mean to cash a promise?
2618/2: { A promise's bearer CAN, by announcement, cash the promise, provided
that any conditions for cashing it specified by its text are unambiguously met.
By doing so, e acts on the creator of the promise's behalf, causing the creator
to act as if e published the promise's text, and destroys the promise. }
# Can a promise create a promise?
Promise creation is permitted by 2618/2: { A consenting player CAN, by
announcement, grant a specified entity a promise, specifying its text and
becoming its creator. }
G. is a player, and 2519/2 explicitly states that a player consents to an
action if { the action is taken as part of a promise which e created }; no
issues there. What about the "by announcement, specifying its text" bit? If G.
just published the text of the promise, it'd certainly qualify, so "act[ing] as
if e published the promise's text" should work fine.
No issues here.
# Can a promise cash a promise?
Sure, it's by announcement. We've already seen that works.
# Wait, can you cash your own promises at all?
This one's interesting. 2466/2 only gives the phrase "acting on behalf" any
special meaning when it's on behalf of another person, so "acting on one's own
behalf" doesn't really have any special meaning or significance. By a
plain-English reading, acting on one's behalf to do something is just doing the
thing. So I think this works just fine.
# What about that conditional in the promise?
Good question, heading. Agora's support for conditional actions is famously not
specified by the rules, only precedent, rooting from 478/38's definition of
performing an action "by announcement" as "unambiguously and clearly specifying
the action and announcing that e performs it". It's widely held that, at some
point, a conditional can become too difficult to resolve to meet the
"unambiguously and clearly" standard. Do we run into that here?
We generally permit conditionals based on ambiguous game state, for converging
the game state and such. This might fall into that category. But this doesn't
really matter, as I shall show below.
Let's say there exists some point at which the state of G.'s vote is so
ambiguous that it can no longer be used as a valid conditional. Where would
that point be? Certainly not after the first flip—at that point, it's just been
flipped once, so it's still perfectly clear. It remains clear after the second
flip, or the third flip, or indeed any finite number of flips; the ambiguity
arises only after an infinite (or arbitrarily large) number of flips back and
forth. Even if the conditional stopped functioning at that point, the state of
G's vote is already unknown. After all, if it was known, the conditional
wouldn't be ambiguous!
So even if the conditional stops functioning at some point, it would only be
after the damage was already done.
# So what's the verdict?
I find PARADOXICAL. Congratulations, G.!
By the way, there was some discussion about my draft ruling on the Discord that
might be worth reading; it's available starting at [0] or in the next Discord
digest.
[0]:
https://discord.com/channels/724077429412331560/724079019578097684/82291900393193472
Gaelan
