Michael Slone wrote: >I don't intend that supermajority voting vanish. So we'll still need a mutability index, or something equivalent, you just don't want it tied to precedence.
There's a basic problem with this. Consider the whole class of systems where precedence is a partial ordering among rules and proposal adoption procedures differ only in the degree of supermajority required. A rational proposal author may be presumed to use the easiest type of proposal adoption that can achieve the level of precedence required. If there is anywhere that a higher precedence can be achieved using a lower supermajority, then proposal authors will use this option, effectively making the higher supermajority (for less precedence) option drop out of the system. The system's effects, taking that phenomenon into account, always yield a strictly non-decreasing supermajority requirement along any chain of increasing precedence. There are more exotic classes of system to consider, which may be what you're looking for. > It might be interesting >to see what would happens were the rules to pretend they contain no >contradictions. Suber's initial ruleset actually works pretty much this way. There are several clauses along the lines of "if not prevented by other rules". We've lost a lot of that cooperative flavour in the current ruleset. With "instrument mechanics" I tried to bring some of it back, among the Power=3 rules. To make the whole ruleset work without contradictions is more difficult. A functionally-equivalent transformation of the current ruleset would have to add hundreds of clauses along the lines of "unless prevented by a rule with Power>1 or Power=1 and Number<2128". *Keeping* the ruleset contradiction-free would entail a reliable procedure to statically detect contradictions, and nullification of proposed rule changes that breach the contradiction-free invariant. We can't do the latter without the former. I'm envisioning writing the entire ruleset in a formal denotational semantics language and automated theorem proving in the pi calculus... we're not going to get there easily. >I don't have a proposal of my own here, There's a refinement of the present system that occurred to me a while ago. We currently have a total ordering precedence, but we could loosen that to a generalised Boolean algebra. Each rule has a set of (zero or more) precedence flags. If two rules conflict, and one has a proper superset of the other's flags, then it takes precedence. If the two flag sets are equal then fall back on numbering for a tie break. If the two flag sets are neither superset nor subset of each other, then there is no precedence between the rules, and the contradiction stands. Not sure what we should do in the no-precedence case, though. Just have the general idea that it would occur with rules that shouldn't interact at all. -zefram
