Date: Fri, 22 Mar 96 10:04:00 PST From: Clark Nelson Message-ID: To: cowan@ccil.org Subject: Re[2]: grammar.246 X-Mozilla-Status: 0011 Content-Length: 2160 X-From-Space-Date: Mon Mar 25 12:10:59 1996 X-From-Space-Address: - John, First, a discrepancy I missed earlier: the BNF says free is allowed after MOI associated with ME, but the YACC doesn't. In any case, if it were allowed, it would have to be inside the brackets with MOI, or there would be shift/reduce conflicts. Also, I forgot to mention yesterday some more places where free is arguably missing, from a consistency standpoint: KE in gek_sentence KE in tanru_unit_2 KE in operator_2 BO in selbri_6 NUhI VUhO For that matter, is there any logic to the places where free is not allowed for which there is no apparent lexical or grammatic reason? Examples include (considering only areas changed recently) after LE/LA, NOI, GOI; also after LI, MAhO, NAhU, SOI (although I wouldn't advocate adding free to these on their own merit; actually, I wouldn't advocate allowing free after PEhE either). The real question is, what are the criteria for deciding where free can and can't go. I will go ahead and include the other locations where I can see no syntactic reason for forbidding free, in case you think it's worth pursuing; if not, please ignore the following list: between adjacent CMENE between adjacent NIhO LA LE LI ZIhE GOI NOI between GOhA and RAhO ME MOI with ME NUhA JAI BE BEI VEI FUhA PEhO MAhO NAhU NIhE MOhE JOhI FIhO SOI XI after COI [NAI] One other quick question, while I'm asking. I was initially surprised, in the grammar for gek_sentence, that tag was not optional with KE. Then it occurred to me that perhaps there is no purpose for KE other than attaching a tag. However, I'm not really convinced -- although I'm having a hard time coming up with an example, I suspect that there may be a case where it is necessary simply to bracket a gek_sentence with KE/KEhE. Even if not, this is effectively the only place in the language where a tag is not optional. Comments? -- Clark Nelson clark_nelson@ccm.jf.intel.com