Return-Path: Received: from SEGATE.SUNET.SE by xiron.pc.helsinki.fi with smtp (Linux Smail3.1.28.1 #1) id m0t6ilA-0000ZOC; Sat, 21 Oct 95 20:35 EET Message-Id: Received: from listmail.sunet.se by SEGATE.SUNET.SE (LSMTP for OpenVMS v1.0a) with SMTP id CB3D4CD9 ; Sat, 21 Oct 1995 19:35:08 +0100 Date: Sat, 21 Oct 1995 14:33:12 -0400 Reply-To: Logical Language Group Sender: Lojban list From: Logical Language Group Subject: Re: Incredible! X-To: MarkLVines@EWORLD.COM X-cc: lojban@cuvmb.cc.columbia.edu To: Veijo Vilva Content-Length: 11606 Lines: 228 >Can the rafsi defects & problems still be corrected? In a word, no. Even if it were accepted that your concerns were valid, any drastic design changes are not permitted at this point. Indeed, even minor changes are difficult, except in some very small areas. We are very close to the 5 year baseline, wherein the language will be frozen as to prescription for 5 years or more (informal and experimental usages will perhaps be developed, but they will by intention not be officially recognized during a baseline). A far more minor effort at making the rafsi attune to usage went through 2 years ago, and it took the commitment that the rafsi list would be baselined to get support for that. Several people were involved in that effort, and concerns such as yours were discussed as part of the debate. >If they are not corrected, can Lojban serve all its various functions? In short, your issues have no effect on Lojban's ability to serve its functions. The only issue is one of learnability, and the tradeoff that was made 2 years ago was that the problems in learnability of "illogical" rafsi were small, and in any case no worse than the cost of relearning foir people who already had used and learned some of the original rafsi. As to whether people are satisfied by the status quo - in general the "old-timers" are satisfied. New people tend to be dissatisfied when they run into something that seems illogical. Sometimes the concerns are valid; in many cases it is a problem that we have not considered, or considered only cursorily. But for the most part, such problems are in semantics and not in the basic language design, which is sound. AS to whether the language design is "elegant", that appears to be a subjective decision. I think that the rafsi system IS elegant, because I value compromises, and am also aware of the flaws with many alternatives. But explaining the "whys" of each design decision to each new person isn't really practical. Someday, we'll have to write a book saying why each design element came to be the way it was, and people could decide for themselves. But even to do this, someone considering the elegance of the design has to bear in mind that all decisions must be made in context. In any design effort you make some basic decisions first, and these will constrain all other decisions. The fine points of assigning rafsi to meanings is a relative "small" decision, that does NOT affect the fundamental principles of the language design, and hence must be constrained by those fundamental princioples. Within those constraints, I contend that the current tradeoff is quite elegant, and furthermore, close to optimal. Not everyone agrees with me on this. But no one has proposed any alternatives that even come close to meeting the many criteria that are applicable, and even if they seemed to do so, we would be unlikely to consider the idea seriously because of the lateness in the design phase. IN general, I have to admit that in anything dealing with fundamental design questions, the first reaction to a newcomer's proposal for change, or even a complaint, is to dismiss it on the assumption that the new person doesn't have enough knowledge of the language to even understand all the unwritten considerations that go into the design. I'm giving your idea a more serious response to most primarily because today I am avoiding some other work that I should be doing instead, and a lot of not-so-newcomers know that this is something that I should NOT be doing. In general, the people trying to produce the books no longer have time to give serious consideration to any design ideas, except those that come up in the course of book-writing, and those that are raised insistently, and survive debate on Lojban List to gain support from at least one of the not-so-newcomers who has been more actively involved in the Lojban List discussion. All this weasel-wording aside, I'm going to respond to your specific ideas. >The rafsi suffer from several very serious defects & problems. Here is a >partial list, with only a few examples of each: >* cmavo & rafsi identical in form > but unrelated in meaning: > da'a (all except), da'a => damba (fight); > mo'i (space motion), mo'i => morji (remember). > >(That was the only problem I'd recognized before, & I'd underestimated >its seriousness. It's really a form of homophone ambiguity -- & Lojban >is supposed to be free of such ambiguity.) This is not an ambiguity at all, much less one of homophones. That is because cmavo are words, and rafsi are not. The cmavo cannot be used as combining forms into lujvo, and the rafsi cannot be used EXCEPT as combining forms. The morphology of the language allows unambiguous resolution of combining forms from separate words, such that it has NO CARE AT ALL about the semantic definition of the words, and only the slightest consideration of the YACC grammar (and that is because a couple of cmavo serve metalinguistically to override some of the basic rules: specifically the non-Lojban text quotes la'o and zoi - within those quotes words need not follow the Lojban morphology and therefore those constructs must be idnetified and filtered out before other grammar processing) Perhaps it would be better to understand that the rafsi for "fight" is NOT "da'a", but rather "-da'a" or "-da'a-" or "-da'a". >* cmavo & rafsi identical in meaning > but different in form: > da'a (all except), daz => da'a (all except); > mo'i (space motion), mov => mo'i (space motion). > >(As you can see, the same forms & meanings are sometimes involved in both >type 1 & type 2 defects. What a mess! You'd think that, if the cmavo >da'a really needed a rafsi, it could have one identical to itself, but >no! It has daz instead, while the rafsi da'a belongs to the gismu damba. > Logical? Not!) Actually quite logical based on considerations of language. That someone MIGHT weant to make a lujvo involving "all-except" justifies giving it a rafsi. But with few exceptions, the cmavo that COULD be used in lujvo are in fact seldom used. manwhile, as you point out in other points, there are plenty of words that ARE usable in lujvo, and of them, at least one - damba - is permitted under the rafsi selection rules to have the rafsi form "da'a". So the tradeoff on this particular rafsi somes down to: will "damba" be desired for use in lujvo more or less often than the cmavo "da'a". Since only one of the two can have that particular rafsi, and neither has a logical (mch less rule-permitted) basis for having any other word-final short -rafsi form, then only one of the two concepts will get such a rafsi. In actualy debate and usage, damba WAS used more, so damba got the rafsi. >* gismu whose meanings ought to be "affixable" > but which lack rafsi: > matra (motor); vidni (video); risna (heart). > >(There are, by my current estimate, some 238 gismu which lack rafsi; but >not all of them have meanings which ought to be "affixable.") At this point, we see one fundamentalgap in your understanding. EVERY gismu has at least one word-final rafsi, and at least one non-word-final rafsi. These are not listed in tables because they are implicit in the gismu word-form themselves. The rafsi for "matra" are "matr-y-" and "-matra" respectively. > 4 >* gismu whose meanings need not be "affixable" > but which do have rafsi: > smo => smoka (sock); bik => bikla (whip); > rig => rigni (disgusting). And this point at least makes me supsicious of another understanding that is missing. On what basis are you saying that these gismu have meanings that need not be affixable, whereas the ones in point 3 "ought" to ba affixable. That you can think of a lujvo for one, and cannot think of a lujvo for the other is not a very strong criterion - you are a single individual, and an English native speaker. Perhaps someone from a different language background or culture would feel just the opposite of you and would not see any reason for a lujvo for risna, and plenty of reason for bikla. NO analysis of word meanings can provide a culture-free basis for deciding which of two words needs a rafsi more than another. The only practical basis is to look at actual usage and proposed usages and see which forms competing for a single rafsi ARE most useful. Thus, under the current design, the only short forms permissible to "vidni" are "vid-", "vin-", and "-vi'i". You would thus have to argue that vidni would be substantially more usefully assigned toone of these rafsi than the current holder of those assignments. The existing rafsi assignments were made on the basis of actual usage and proposed usages, and even 2 years ago it would have taken SEVERAL more lujvo based on vidni to get it to displace one of the others. > 5 >* gismu whose meanings ought to be "suffixable" > but which have only non-suffixable CVC rafsi: > meb => mebri (brow); run => rutni (artifact); > tab => tabno (carbon). This is dealt with by the same argument. But I am particularly curious about one. In what way is "mebri" particularly appropriate to ahve a final-position rafsi, while smoka does not justify any rafsi at all. (notwithstanding the fact that mebri does have the non-shortened "suffix" form "-mebri". > 6 >* rafsi with parallel forms but which are derived > from dissimilar valsi: > cme => cmene (name); zme => guzme (melon); > gu'a => gunka (work); bu'a => bruna (brother). > > 7 >* gismu with parallel forms but which are contracted > into dissimilar rafsi: > cabna (now) <= cab; zabna (favorable) <= zan, za'a; > senci (sneeze) <= sec; denci (tooth) <= den, de'i. You are taking a very narrow view of "similar" and "dissimilar". Your cases in 7 differ only by one letter in the initial position. Fine. But how about "carna" and "cabra" which similarly differ only by one letter from "cabna" Do you consider them to be just as "parallel"? But it turns out that one of these - "cabra" has a CCV form possible - "bra", while the other two do not. Thus among those three, they really AREN'T exactly parallel in structure. By the standards of Lojban rafsi-making, the words in question DO have similar structures as we define similar. Specifically a CVC form rafsi coming from a gismu of form C1V1C2C3V2 can use C1V1C2 or C1V1C3 and one of form C1C2V1C3V2 can use C1V1C2 or C2V1C3 Similar constrainst apply to CCV and CVV form rafsi. It happens that in 6 you tried mixing CCVCV gismu with CVCCV gismu, and of course at the surface ALL such gismu will appear to be of "dissimilar" forms. Why should either of the two forms take precedence for assignment of a CVC rafsi? >In fact, I found so many rafsi problems that I now favor scrapping the >whole rafsi-lujvo system & starting over. At the very least, we should >discard those 295 rafsi, each of which is identical in form to some cmavo >but unrelated to it in meaning. Would you folks agree with that? Given your example of problems that you find with the system, it seems likely that whatyou object to is NOT the rafsi system itself, but the methods of rafsi assignment - a far less basic issue (albeit one not open to change these days). I strongly suspect that the lack of elegance and beauty that And sees in the rafsi system is far more basic than the level that you are seeing "ugly". I'll let him say what his problems are, if he wishes, but having debated with him before on the issue, I suspect that our differences are based more on different aesthetic assumptions, and are hence subjective. lojbab