> I think G.’s judgement in that CFJ is correct (if I understand it right).
>
> G.’s decision says that when a report self-ratifies, it does not change
> anything about the gamestate immediately prior to the publication of the
> report. That makes sense to me. However, self-ratification CAN retroactively
> change the EFFECTIVENESS of actions taken during the time between the
> report’s publication and its later self-ratification—I think that is
> consistent with G.’s decision.
>
> It is also demonstrated by the failure of my recent attempted PARADOX scam.
>
> I had a PARADOX scam that hinged on an ambiguity about who the Prime Minister
> was—me or ATMunn. If intents remained broken then it was one of us; if
> intents were retroactively fixed then it was the other one of us. I tried to
> create a PARADOX by using the Manifesto power to distribute a proposal that
> would fix intents retroactively. Trying to resolve that Agoran decision would
> create an infinite loop where the identity of the PM would cycle around and
> the EFFECTIVENESS of the distribution-by-Manifesto cycled too.
>
> It was foiled by the retroactive effect of a self-ratified ADoP report.
>
> At the time I issued the attempted Manifesto, the ADoP report had been
> published but had NOT self-ratified. If I had CoE’d it before ratification,
> then the identity of the PM would be ambiguous and so the Manifesto attempt
> would have had ambiguous EFFECTIVENESS. But I missed the ADoP report and it
> later self ratified—causing the Manifesto attempt to be unambiguously
> INEFFECTIVE. Which was very, very frustrating.
>
> The facts are recounted in detail in CFJs 3722, 3723, 3724, and 3725.
>
> I think that conclusively shows that the attempted blot levy was INEFFECTIVE.
My complaint with this argument is the following.
When R1551 says:
"...to what [the gamestate] would be if, at the time the ratified
document was published, the gamestate had been minimally modified..."
this is talking about a hypothetical sequence of events, and yes,
during that hypothetical sequence of events, certain actions that we
previously considered to be EFFECTIVE are hypothetically considered to
be INEFFECTIVE, and vice versa. But that does not imply that the
hypothetical past is itself part of the gamestate.
If the history of EFFECTIVE and INEFFECTIVE actions is part of the
gamestate, then I agree that the effect of R1551 is also to change
that history. But it does not need to be part of the gamestate in
order for R1551 to have the effects we generally rely on it for.
When we fixed intents, this was my interpretation of what happened:
* Proposal 8164 refers to "what [the gamestate] would have beeen..."
in a similar manner to R1551.
* In that hypothetical alternate history, the EFFECTIVENESS of actions
is different from reality, and therefore the gamestate (switches,
ownership of entities, probably other things...) ends up different.
* Proposal 8164 changes the gamestate --- all those switches,
ownerships, etc --- to how it ended up in that hypothetical alternate
history.
I don't think CFJs 3722, 3723, 3724 or 3725 directly address whether
or not the EFFECTIVENESS (or POSSIBILITY) of past actions is changed
after ratification:
* I think 3722 was resolved without reference to ratification.
* 3723 is present-tense; it's about whether the current (at time of
CFJ) gamestate allows an action. It does discuss a hypothetical game
history, but does not assert that that hypothetical history actually
happened.
* 3724 did not involve ratification.
* 3722 was about the EFFECTIVENESS of an action being affected by
ratification. However, I think in that case the document ratified
before the action was even attempted, so there's no question of
changing the past.
So, it's still unclear to me whether the past is part of the
gamestate. I see no reason why it should be. Here's a pitch explaining
that, at least, this viewpoint is self-consistent:
Let's define R(T, X) to mean: at time T, the Rules said that X was
true. In this definition, I intend R(T, X) to be fundamentally a
statement about reality, not directly governed by the Rules; it means
that a diligent player familiar with the rules of Agora, sitting down
at time T and carefully studying the situation, would conclude that X
is true at that time.
When a CFJ about past effectiveness is called, in reality, the player
who's being the judge presumably sits down and tries to work out:
R(now, [at the time the CFJ was called, action A was EFFECTIVE]). We
have to wrap that in R(...) because "EFFECTIVE" doesn't really mean
anything outside the Rules.
But, "at the time..." does mean something outside the Rules, so I
think it's reasonable for the judge's first step to be to unwrap that
as the equivalent statement: R(time CFJ was called*, [action A was
EFFECTIVE]). If the Rules say that's wrong, then of course it's wrong,
but if the Rules are silent on the matter it seems reasonable.
*More pedantically: the judge will first have to find some time T in
the past such that R(now, [T is the time the CFJ was called]), and
then judge R(T, [action A was EFFECTIVE]).
