On Thu, 14 Jan 2010, ais523 wrote:
> Valid point, I'm pretty convinced that ratification is broken, now.
> (Possibly by being too ambiguous to have an effect.) I still don't think
> the general concept of ratification is too weak to provide a mechanism
> around 1698; but the specific implementation we have at the moment may
> be.
I agree. I think we tried to be parsimonious and general by using the
phrase "the gamestate is minimally modified..." but it leaves far too
much open. Does it modify peoples votes or just the fact of adoption?
Does it modify state or whether events happened? (either one could be
the "minimal" modification depending on circumstances). And it's very
unclear, when ratification happens, what event caused the values to
change.
I think we'd actually gain something by admitting true retroactive
action rather than asking us to believe events that didn't happen.
Eg, "when a document ratifies, the ratified quantites are retroactively
set to be what they are in the report at the time of the report, and
the time of the change (if any change was made) is considered to be
the time the report was published."
So if I have 5 points, and my score is (accidentally) self-ratified
to be 10 points, we can say that "my score was changed from 5 to 10
at the instant the report was published" which we can't actually say
now. (This still leaves us open to occasional retroactive action
questions or possible paradoxes, but I think it's cleaner and anyway
a new source of paradoxes is part of the fun). We should also add
specific language for when ratification sets something to an impossible-
to-hold value.
We also need to limit ratification to gamestate quantities and not
"the gamestate as a whole" or "any document" including events that
didn't happen. So you shouldn't be able to ratify the fact that a
message (e.g. a vote) was posted at a certain time when it wasn't, but
you could ratify a vote of A when the actual vote was B, and this
would mean that the publication of the report changed the vote
from A to B (as long as it has precedence over R2034 of course).
The following rule used to exist and was useful; it distinguished
things that could be ratified from facts that couldn't be changed
(e.g. events). [The term 'properties' is unfortunate because it
became confused with asset-type property, but it's a good concept].
It would mean re-writing some critical things as states rather than
events to allow ratification (e.g. the "adoption state" of a proposal
indicating whether it passed or not) but that's better than ratifying
actions that didn't happen.
Rule 1011/4 (Power=2)
Definition of Nomic Property
A Nomic Property is any property of any entity the value of
which is defined by the Rules. Other Rules may define
procedures by which the value of a Nomic Property may be
changed.
-G.
