In general, I'd tend to agree with G.. - No *platonic* failures need apply unless a truly verifiable form of randomness is used; in lieu of that, you can always cheat just by lying about a roll, so bad randomness would always be a SHALL violation.
- The basic requirement is that you SHALL make a reasonable attempt to ensure no irregularity occurs (or at least minimize irregularity) in the rolling process. An irregularity exists if the mapping of dice rolls performed to hypothetical final gamestate result, considering all players' actions during the relevant time period, does not have the distribution expected by the rules. For a variety of reasons, an irregularity could still occur despite everyone involved making such an attempt, but it's unlikely enough that we don't really have to worry about it. - An additional requirement is that everyone SHALL avoid making decisions regarding which they have privileged information (i.e. as yet unannounced rolls) that would give em an unfair advantage. That includes a decision not to act. - The legality of most of ais523's hypotheticals can be easily resolved under these metrics, with some special cases: > C. An officer selects a random result, then waits a week. The > following week, e communicates it to Agora. If they actually looked at the result (as opposed to blindly saving it to a file) and could stand to gain in any way during the next week based on this knowledge, they would breach the second requirement. Otherwise, it wouldn't breach either of my requirements, precisely because it would be hard to demonstrate any harm in practice. Because there is no real justification for queueing up randomness this way, you could add an additional requirement to communicate results with haste, but it wouldn't be essential. > N. An office is empty. A nonplayer selects a random result. > Disliking the result, e does not register (leading to no visible > evidence of any of this on the mailing lists). If there is no visible evidence, then it's undetectable and therefore unavoidable, but probably not a big enough problem in practice to justify verifiable randomness by itself. Supposing e instead flaunted the fact for some reason... > R. An officer selects a random result. Disliking it, e deregisters. If e happens to announce the precise nature of what e dislikes, it would be legal and obligatory for a replacement officer to ensure the final result ends up that way (re-rolling anything that was not disclosed), to minimize the irregularity. In both cases, the person would violate the requirement, just with punishment being bad form due to their non-playerhood. If it actually became an issue, it would likely be due to people with the habit of ping-ponging playerhood, and we could punish them the next time they registered. Under the old R101, in the second case doing so would arguably be a violation of the right to deregister (if they planned to deregister given a bad result, the roll itself could arguably be punished, but good luck proving that), but that clause no longer exists. In practice, most likely such fits of pique would be one-off cases not really worth punishing.