From - Mon Mar 10 10:01:43 1997 Reply-To: Logical Language Group Date: Mon Mar 10 10:01:43 1997 Sender: Lojban list From: Logical Language Group Subject: terminator after JOI To: geoffreyhacker@msn.com, lojban@cuvmb.cc.columbia.edu X-Mozilla-Status: 0001 Content-Length: 3751 Message-ID: >I object to having to put what is normally an elidable terminator, "ku", in >between "le nanmu" and "joi le ninmu" just because the parser is too stupid >to realise that if a "le" follows the "joi" then "joi" must be connecting two >sumti, not two tanru components. Surely it cannot be difficult to write a >parsing program using the algorithm: > >1. If a descriptor follows the JOI cmavo, then the cmavo is joining two >sumti. 2. If a brivla follows the JOI cmavo, then the cmavo is joining two >components of a tanru. > >and so on, for all the other grammatical things that can follow a "joi". >There is no reason why a parsing error should have to result in reading what >is, IN PRINCIPLE, an unambiguous phrase. So I will just have to formally >object to --More-- >the Lojban formal grammar in this instance. People should be aided, not >restrained, by machines. Your objection is noted, and you aren't the first to do so. Si nce we started with the requirement that the language be LALR1 (i.e. YACCable), this is NOt a "parsing error" but a "design feature". No one has proposed a way to make this work using only single token lookahead, and it may or ma y not be possible. If we had early on had a tool that supported double lookahead, this might not be needed. I should note that JOI itself is kind of an aftertthought in the language design. JCB had only one such non-logical connective and he arbitra didn't look into the semantic nuances that might require it being used both for connecting sumti and descriptions, so he has that connective either in one or the other connective group corresponding to A and JA. We recogn ized the need for multiple kinds of such connectives, and moved them out to a separate selma'o, which alone among connectives can be used EITHER where an E is used OR where a JE is used. The alternative was to use two cmavo for each connective, one for each selma'o. The wide range of places where JOI can be used makes it darn near impossible to guarantee that a given usage is unambiguous without the terminator. No one has come up with a countering ambiguous usage involving JOI followed by LE, but that does not mean that none exists. Even if one existed, this is not the only place where we have had to make an adjustment because of a lack of two-token lookahead (there are several things in the lexer rather than in the regular grammar because of this weakness). IN any event, the handicap of requiring a terminator once in a while where it might seem extraneous, was for us a relatively low priority. It may be aesthetically displeasing to some (and your objection in the final analysis is probably aesthetic), but it works. Including an extra, possibly unnecessary word, is consistent with Zipf's Law design pronciples for the language if we presume a higher usefulness for logical connection than for nonlogical connection, so by that reasoning the extra terminator could even be seen as desirable. (After all, it is a reasonable assumption that a logical language will make it easier to say something logical than something non-logical, given a choice). Extra terminators also provide extra redundancy in speech. But I still note your objection. Since the language is baselined, there will of course be no consideration for changing it. lojbab ---- lojbab lojbab@access.digex.net Bob LeChevalier, President, The Logical Language Group, Inc. 2904 Beau Lane, Fairfax VA 22031-1303 USA 703-385-0273 Artificial language Loglan/Lojban: ftp.access.digex.net /pub/access/lojbab or see Lojban WWW Server: href="http://xiron.pc.helsinki.fi/lojban/"