I thought about this earlier. The problem I have with it is that there's no time at which the switch would fail to have a possible value. If the specified value is invalid, you don't create a proposal that would have an invalid value apart from that provision, you just fail, at least under my theory. The problem isn't that you're not specifying that you want the extra avocado; you're requesting to have your sandwich with an impossible feature, like asking for the sandwich with elephant egg omelet topping. At that point, the waiter tells you you're insane and you're refused service, or at the very least politely asked to make a valid order.
-Aris On Mon, Jul 1, 2019 at 5:19 PM Kerim Aydin <ke...@uw.edu> wrote: > > > I'm actually coming round to D. Margaux's result now, but for a bit of a > different, more narrow reason. R2162(Switches) says: > > If an > > instance of a switch would otherwise fail to have a possible > > value, it comes to have its default value. > > If you created a proposal with AI=0.5, it would absolutely fail to have > a possible value - therefore it goes to default as the proposal is > created. And I think this pretty strongly maps onto the concept of > "default" and "optional". Just in common language, if you order > off a menu and don't specify that you want the extra avocado, they > don't say "you didn't say which options you wanted, no food for you", > they serve up the default. > > On 7/1/2019 3:31 PM, Aris Merchant wrote: > > I’d propose a different theory. Mine is cleaner and simpler, but I’m not > > entirely sure which is actually better. Barring someone is a separate > > action from calling the CFJ; it just has to be done in the same message. By > > contrast, it’s a tad hard to argue that specifying the AI of a proposal is > > somehow separately from specifying the rest of its properties. So it makes > > logical sense that barring should be able to succeed or fail atomically, > > whereas specifying a proposal’s AI shouldn’t. G., any opinion on which of > > these explanations better fits with Agoran practice? > > > > -Aris > > > > On Mon, Jul 1, 2019 at 11:44 AM Jason Cobb <jason.e.c...@gmail.com> wrote: > > > >> IANAAL (I am not an Agora lawyer). > >> > >> I think that a key difference between those two scenarios is whether or > >> not the invalid action affects the gamestate. For instance, the AI of a > >> proposal is a key part of the proposal's identity, it will affect > >> whether or not it gets adopted, what it can do, etc. Changing the AI of > >> a proposal makes the consequences of the proposal vastly different, so > >> it makes sense not to do that implicitly. While for the CFJ scenario, > >> the difference between the barring succeeding and failing is nothing - > >> if the barring succeeded, then the non-person couldn't judge anyway, so > >> no difference to the gamestate than if it failed. > >> > >> Jason Cobb > >> > >> On 7/1/19 11:16 AM, Kerim Aydin wrote: > >>> > >>> On 6/30/2019 11:32 PM, D. Margaux wrote: > >>>> If a player does all that and also specifies that AI=e, I don't see > >>>> why that makes the CAN clause fail. > >>> > >>> It's impossible to create a Proposal with AI=0.5. If I say "I create the > >>> following proposal with AI=0.5" it's equally reasonably to say "no you > >>> didn't, that's IMPOSSIBLE to do and it fails" as it is to say "you got > >>> that > >>> half right - you made a proposal but it's a different (default) AI." > >>> With those being equally reasonable in a vacuum (IMO), we've tended in > >>> interpretation to err on the side of complete failure, for practical > >>> reasons. It's so much easier to simply say "whoops, no action, try > >>> again" > >>> then to say "you did it half-right and now we have to clean up the > >>> mess of > >>> the half-correct proposal." > >>> > >>> I'm not sure if there's a general principle here. For some things (like > >>> paying fees) we really stick with complete failure - e.g. if you try > >>> to pay > >>> a fee and only have 1/2 the number of coins, we don't say "you paid > >>> half and > >>> then fail to perform the action" - there's an implicit "if this fee > >>> can't be > >>> paid in full, it fails". > >>> > >>> On the other hand, let's say you call a CFJ and try to bar a > >>> non-person from > >>> judging for some silly reason. That "bar" fails, but I'm guessing > >>> we'd just > >>> let the CFJ go through - because the CFJ-calling and the barring are > >>> somewhat weakly connected. > >>> > >>> This is all to say - this seems like the sort of thing were some general > >>> principles/tests along the continuum of success/failure are worth > >>> figuring > >>> out. > >>
