Return-Path: <@SEGATE.SUNET.SE:LOJBAN@CUVMB.BITNET> Received: from SEGATE.SUNET.SE by xiron.pc.helsinki.fi with smtp (Linux Smail3.1.28.1 #1) id m0s6Uri-0009acC; Wed, 3 May 95 06:12 EET DST Message-Id: Received: from segate.sunet.se by SEGATE.SUNET.SE (LSMTP for OpenVMS v0.1a) with SMTP id FA57AB48 ; Wed, 3 May 1995 5:12:40 +0100 Date: Tue, 2 May 1995 20:10:49 -0700 Reply-To: "John E. Clifford" Sender: Lojban list From: "John E. Clifford" Subject: proposals X-To: lojban list To: Veijo Vilva Content-Length: 11275 Lines: 162 On "any" still once more, if all that is lacking is conciseness, I should note that _pa_ is shorter than "any" by any measure, and _CVhV_ro_, while longer, is about the same relative length, given that lojban expressions tend to be longer than English. Now, what is lacking -- except perhaps the will to use logic effectively? On xorxes's example. When I said that I did not understand them, I meant just that: they were examples without adequate explanation of what was involved, what was being changed to what. In light of the further discussion with lojbab, I now see a bit more of what is involved and am less enthusiastic, at least for the _du'i_ case. that seems to involve moving an expression from an adverbial to a conjunctive role and that kind of move, initiated by a single word but then encompassing a whole lexeme, seems to me to be near the heart of a number of odd changes in recent years, changes that make it hard for me to recognize much of current lojban structure from a historical perspective. Since the sentence involved does seem to contain a logical conjunction (actually of three sentences) and the addition is only an _e_, I would vote to stick with the present form unless a lot more study of the consequences of the shift show some advantages to it. On leapers and (to coin a phrase) pointers. A leaper does indeed change the prenex, placing the tagged quantifier at the head. The other term, the pointer, just indicates that the tagged sumti is to be interpreted from outside the opaque context in which it lies; it does not -- if it is a quantifier -- actually move it outside (although I suppose it has that effect eventually). More importantly for the introduction of the pointer was that if allows nonquantifier terms in opaque contexts to have external reference and (the specific xorxes problem for which it was introduced) thus to be used after opaque-making predicates without the subject-raising flag and, thus, to be treated as transparent. It is a minor point, but a useful one, as xorxes kept insisting last fall. On lambda. I always assumed -- and those who have used the forms have practiced -- that insertion into an abstraction was in order from the left into free spaces. (This is the rule in lambda too, except that rarely are the spaces left free to apply it, it is how modified lambda works, however.) Thus, _le_ka_dunda_ is the property of someone giving something to someone and applies first of all to a giver and then to ordered pairs and finally to ordered triples (this is not exactly how it works in lambda but the end result is the same). To get the recipient, then, we have either to convert the predicate _te_dunda_ (isn't it?) or fill in the unwanted gaps explicitly _le_ka_da_dunda_de_ (scope only within the abstraction). We old- timers know that we can get any ordering of terms to a predicate using only members of SE, but for this purpose, we can achieve the result we want by choosing the order of the terms to be inserted and by blocking out uninteresting places. (The possi- bility of odd orders -- including identifying initially different places -- is the main use of full lambda, which presupposes that the terms to be inserted are in a given order already, a condi- tion we need not worry about). The biggest problem with making full use of this insertion procedure is remembering how to do order n-tuples. "John loves Marry more than Harry Sally" is something along the line of la djan ? la maris zmadu la xeris ? la selis le ka prami but "John loves Mary more than (he does) Sally" is simply la maris zmadu la selis le ka la djan prami (or se prami la djan). And so on. It was nice to see that xorxes also see this pattern and seems to approve. Neither the dakau nor the ke'a plan seems to offer any advantages over what is already in place, if implicit. (But I have to admit that the possibility of trying to get reflexives in does suggest one use for the fuller lambda form; repeating the term involved seems unsightly.) Sorry about the tensor talk. Over the years, the two factors in mathematical tensors and vectors as they apply to tense (and location) in language have been separated and given the names "vector" (direction) and "tensor" (length), probably to the dismay of the mathematically inclined person who started using them in the first place. Lojban now has these factors sorted out into two lexemes for each dimension realm. No problems seem to arise with the vector forms (at least none now under discussion). The tensor forms do seem problematic, at least to xorxes (and perhaps others, since they do not seem to use them much). Nor are there any mentioned problems with the tensor forms in tense position, the indicate a vaguely defined distance in an unspecified direction (apparently, in the case of the temporal tensors at least, the direction is taken as being important enough to infer from context), just as the vectors are pointing in directions of unspecified length and the mixed forms give both length and direction. In all of this, the head of vector/tensor is the event of the bridi, the tail is a contextu- ally specified origin. For vector forms as sumti tcita, the head is still the event of the bridi and the tail is the origin de- fined by the tagged sumti, corresponding to the hypothetical role of the hypothetical contextual axis-register in the form underly- ing the tense-location form (and the other adverbial forms). The same seems to be true for the mixed vector-tensor forms. So, by parity of reasoning, the pure tensor forms should be defining a displacement without direction from the named event/location. And the spatial tensors seemed to be used in this way: "near to", "around", "a ways away from", roughly. So, why don't the temporal ones mean the same thing, roughly various approximations to the time of the event named? While the uncertainty about direction is more troublesome in the temporal cases than in the spatial, the forms still have some uses and should be kept. Two related issues come off of this. 1) Xorxes claims that the combination of the 0-vector and a tensor does not make sense. I am inclined to think it is the moist useful form of the vector- less tense form. The mathematical 0 vector is in reality (that is, outside mathematics) not a 0 at all, but an indefinite range in the appropriate dimension(s). as the old saying goes, "When- ever I say 'now' it is already then" and "Only I can stand here": by the time I utter the flag of the instant, the instant has passed, for someone to come to my here is for them to be coexten- sive with me. So, in fact, we use "here" and "now" (and "at". and so on) in very unmathematical ways. But we sometimes like to say just how far off we are, how broadly we are using the terms and this fuzz factor then needs to be conveyed. In English we have a lot of "about" locutions and the like, e.g., "nowadays" for a moderately extended now (a few years rather than a few geologic eras). I think that this fuzz factor is just what tensors do on 0-vectors. Notice that _vi_ alone (which I would take on the usual grounds as being _bu'uvi_) is just a moderately broader _bu'u_, allowing something other than perfect coinci- dence, though expecting some propinquity. In the same way, I would take _zi_ (i.e., _cazi_) as being "now in a slightly ex- tended sense," a close as makes no nevermind to the time. The more remote forms of course allow more difference, though still irrelevant for the present purpose: I suppose _(ca)za_ is about "nowadays". 2)The fact still remains that the tensor system is inade- quate (and so, I think, is the spatial vector system -- can we really even box the compass in lojban, let alone lay out a real direction (or full tensor in the mathematical sense) in three - or four - space, as we would need to do to, say, give instruc- tions to an anti-aircraft gunner?) We can -- and once did -- have a tensor form that lays any appropriate metric down: "three days," "seven feet," or whatever is needed. I take it that goran has proposed returning to this, as has xorxes (though, in the latter case, at the cost of roving other items from the internal system). I think that, for clarity, we need two such operators, temporal and spatial, since we may want to specify in both areas and, although most metrics will be distinct, confusion still might arise (lightyears seems to confuse most people, for exam- ple). The other problem is to fit all these pieces together. If we are using a specific tensor but no vector, how do we specify the origin, how, that is, say "thirty yards from the school" or "three days from your birthday". The obvious suggestion here (another cmavo, of course) is to have an origin marker (or two, time and space again) indicating that further specification is to come, for attempting to get origin and metric in the same con- struction is going to tax the grammar again. In fact, we may be able to recycle some of the existing forms for some of these purposes; the _zi_ form seems ideal for the direction unspecified "within three days of your birthday" and the _Cu_ forms might work for "more than n units away" (not quite in line with the framework above but manageable). Of course, if the vector is specified, the tensor does not need help. And what about very specific vectors (only the spatial ones present problems except in really esoteric -- and highly spatialized -- tense logics)? We may have two or even three terms involved depending upon the conceptual framework within which we are operating (xyz or angles or bearings and altitudes or...). Of course, all of these can be reduced to tuples of various sorts but spelling them all out seems more reasonable. Sounds like a place for a commitment to some further study. But the tensor cases can be handled immediately and should be, at least to the level of assigning _xVhV_ forms for the two types.