Discussion: Type System

I completely agree that Lojban has a type system, and I've treated it that way. In fact, my jvajvo rules rely heavily on sumti types. E.g. the infamous {karcykla} (without pruning) becomes

x1 klama x2 boi x3 boi x4 boi x5 noi karce x6 boi x7

because karce1's type is "vehicle", as is klama5's. Matching up sumti types is the basis for my jvojva.

With that said, I would like to start a discussion about the implementation of a type system in Lojban.

Some points I'd like to discuss are:

  1. lo and masses
  2. Type conversions
  3. Sets


--selpa'i

lo and masses

  1. Personally, I do not see why lo should not be able to create masses. Having an agnostic-about-masses article saves you the trouble of always deciding whether or not you really mean individuals or masses. You can still be explicit if you want to, using loi for explicit masses and lo with outer quantifier for explicit individuals. What are your reasons for wanting to disallow a lo-mass? What do you do with lo and inner quantifiers which are masses as well? Are you saying lo mu bakni is not a mass? --selpa'i
    1. Yes, I'm saying that {lo mu bakni} is not a mass. It's some individuals, out of the five contextually relevant cows. Sure, having an agnostic article would be nice, but I don't want it to be {lo}. While we're at it, why not let some article produce some super-vague polymorphic description that can be a set, a mass, a description, and event, a property.... I don't want to overload {lo} which a million meanings. Also, the amount of forethought required to properly use the correct type is almost nonexistant. The only thing that this limits is the usefulness of {gi'e} which one can just replace with {ije} and then the correct sumti, if you reject {jai'a}. — tsani
      1. I don't think individuals/mass/(set) is a case of overload. {lo} does a few simple things, but it doesn't do what LAhE do. LAhE are a seperate thing used to convert sumti types. {lo} only creates sumti. --selpa'i

Type conversions

  1. My main problem with your idea is this: Since every sumti-place has an inherent type, it should not be necessary to explicitly state that the sumti you put in it is of that same type. Take traji3 for example. What's the point of saying {ti traji lo ka cizra kei lo'i dacti poi mi viska ze'a lo mi nunjmive}? traji3 is a set by default, putting in any sumti makes that sumti a set automatically. (traji: x1 is superlative in x2 (ka) among set/range x3) --selpa'i
  2. I should mention that a sumti place can allow for more than one type, often both a dacti and a fasnu, but sometimes different things too. --selpa'i
    1. The sumti places that truly are multi-typed are extremely few, in my opinion. The only notable one is nelci2, which allows for any type from class A. The point behind saying {lo'i} rather than, say, {lo} in traji3, even though we all know that traji3 is a set, is for precision. Sure, you can say that there're implicit LAhE or whatever that's necessary to produce a reasonable result, but that's just not what I believe. Your jvojva rely on the fact that types be consistent and static, but it's okay to just put whatever sumti in whatever place and let the selbri do all the work in making sure that types be consistent. That seems illogical to me. My belief is simple: types are static, and the speaker must ensure that the correct-type sumti go in the correct-type places. If the referent yielded by a description is not of the correct type, then it is necessary to use a cast operator, like LAhE.
      1. I understand your standpoint. But I don't see why this extra effort is worth it. You say the selbri does all the work, but the selbri doesn't turn lo into lo'i, lo does. lo itself can stand for lo'i because lo'i is a member of LE. lo is the father of LE and can stand in for any of them when no extra specificity is required or desired. My main motivation is usually to impose as little restriction as possible. Your proposal makes a vast number of sentences "wrong", akin to a wrong conjugation in a natlang, which to me means a restriction in the way someone is allowed to speak Lojban. You are forcing people to use lo'i and put lu'i/lu'o etc around sumti, which means a lot of extra syllables for no gain other than more explicitness, but not more correctness if you accept that lo is a chameleon. --selpa'i

Sets

  1. You say that simxu1 is a set. I say it can be a set, but can also be a mass. If you limit simxu1 to a set, you make it rather useless. For example, you cannot say {lo simxu be lo ka ce'u tavla ce'u cu klama lo panka} because sets cannot go. Do you agree? If not, what does a set mean to you in the context of Lojban? --selpa'i
    1. I disagree because we have LAhE for that, not to mention that you believe that type can be casted automatically by the selbri. Yeah, simxu1's type is inherently "set", but if you stick a set in a mass place, according to you, it becomes as mass, by implicit LAhE. Sets in Lojban are to me pretty much just like sets in math. They have no properties other than those that they all share with one another, like cardinality. Perhaps getting rid of sets altogether is a good idea, replacing them with masses everywhere (I do say {mi'a simxu..}).
      1. You can't argue for your position by using mine. You do not seem to think that lo simxu is a mass. And under that view, the sentence with klama doesn't work, because sets can't klama, which you agree with as well it seems. Regarding LAhE, if you have such a rigid system, that means you have to use LAhE all over the place to have the right type agreements. Do you think that's worth it? Saying lo'i everytime instead of lo means one extra syllable every time. --selpa'i


Created by selpa'i. Last Modification: Saturday 07 of July, 2012 21:55:48 GMT by selpa'i.