Return-Path: Received: from SEGATE.SUNET.SE by xiron.pc.helsinki.fi with smtp (Linux Smail3.1.28.1 #1) id m0t6ac0-0000ZOC; Sat, 21 Oct 95 11:53 EET Message-Id: Received: from listmail.sunet.se by SEGATE.SUNET.SE (LSMTP for OpenVMS v1.0a) with SMTP id DE44CA46 ; Sat, 21 Oct 1995 10:53:07 +0100 Date: Sat, 21 Oct 1995 05:51:59 -0400 Reply-To: Logical Language Group Sender: Lojban list From: Logical Language Group Subject: Re: lujvo-making X-To: jorge@PHYAST.PITT.EDU X-cc: lojban@cuvmb.cc.columbia.edu To: Veijo Vilva Content-Length: 2022 Lines: 42 >> Because lujvo-space takes precedence over fu'ivla-space, this cannot >> fall apart: "cporba'u". would be a fu'ivla. > >So the rule is that "cporba'u" cannot be a fu'ivla, right? > Correct - that word is exactly of the "slinku'i" form for which the test is named that rejects such words. >I thought that the 'y' between consonants was always allowed, even >between a pair within the same rafsi, because some people might have >problems with some consonant clusters. Was this ever so, or did I just >dream that up? Never so. This is the distinction between the hyphen which is generally a schwa, and the buffer, which must be some other sound and hence is often a higher lax vowel, like English short-i or IPA barred I (the Russian "61" letter). In my careful speech, I tend to make the hyphen a bit more back, and the buffer a bit fronted. The buffer is also a much shorter sound than even the short hyphen-schwa. >Have the rules for fu'ivla ever been worked out? I thought it was >just "anything that's left" after all cmavo, gismu, lujvo and names have >been removed, but that is not enough, because that would allow "cporba'u", >which cannot be any kind of word. > The definition is close to what you suggest - it is anything that in a Lojban speech-stream cannot be taken as a gismu/cmavo/lujvo/name. cporba'u fails because of the Slinkui test, which is that if you insert a CV cmavo in front of it, the result in the speech stream is a lujvo "pa slinku'i" becomes "pas-lin-ku'i". There are several rules that are know that help identify illegal fu'ivla but they are far from intuitve and there has neer been an effort to devise a complete set of rules, which would of course be a set of strings of CVCCVCVCVCVVC etc. that would break up. Nora's morphology algorithm implementation would do this, but even generating all the test cases of, say 12 characters or less, would probably exceed reasonable computation time, unless there was a smart case generator/weeder (I know because I tried %^( ) lojbab