On Sun, 28 Oct 2018, Reuben Staley wrote:
8121 G. 3.0 Retroactive Documents
PRESENT. There are arguments for and against doing this and I am not
convinced which way to go.
This was a result of a discussion between me and G. on a rather subtle
point, where it took a while for me to convince him, so given the proposal
has no comments, I'm not surprised that some voters don't see why it's
needed.
Ratification _simulates_ retroactive modification of the gamestate, but
it's written to achieve this without _actual_ retroactivity, with a
two-level mechanism:
* Find a _minimal_ hypothetical modification to the gamestate that would
have made the document's claims true if they had been applied _at the
time it was published_.
* Then fast-forward from that hypothetical modification to the present,
and in the present change the gamestate to what it would have been if
the hypothetical modification had been done in the past.
This mechanism works well when applied to documents that describe
"concrete" gamestate _at the moment they are published_, because at that
moment the minimal change needed to make them true is generally obvious.
But not all documents happen to be of this kind. Sometimes a CoE results
in a revision where the Officer publishes a new report, but with a note
that it does not include changes since the previous report. And
sometimes, like the hypothetical that started this discussion, someone
attempts to correct an error in the past by publishing a _new_ document
speaking _about that past_, and then ratifying it.
However, with the current wording of ratification rule, this can have
strange effects, because there is now time for _changes_ in game state to
happen between the time the document speaks _about_, and the time it is
_published_:
Time A: G. has 50 coins
Time B: G. uses coins for scam (needs at least 40), massive follow-on game
changes
Time C: Report (or possibly anti-scam document) published, says G. had 10
coins at time A
Time D: Report ratified somehow
By the current Ratification rule, the "minimal change" is calculated at
time C, _after_ the massive game changes caused by the scam.
At that time, is a _smaller_ change to say that the scam _worked_, say by
having G. lose 40 coins at time A, but then regain 30 some time before
time B. Therefore, by the wording of the Ratification Rule, it does _not_
cancel out the scam, as one would otherwise intuitively expect.
G.s proposal adjusts ratification so minimality is calculated at
time A, instead, where again the required change is obvious and has the
intuitive result of cancelling the scam.
Side note: The trouble would be even more severe if G.s scam involves rule
changes, because then the intuitive (but huge) modification at time C
would need to cancel those, _without_ the document saying so, which the
ratification rule _also_ explicitly forbids.
Greetings,
Ørjan.
