From: cbmvax!uunet!bigd.cray.com!dmb
Message-Id: <9104121339.AA02476@bigd.cray.com>
To: Guy Steele <gls@think.com>
Cc: nsn@mullian.ee.mu.oz.au, lojban-list@snark.thyrsus.com
Subject: Re: Elision, or: Nick rides again in jbonai 
In-Reply-To: Your message of "Thu, 11 Apr 91 15:38:36 EDT."
             <9104111938.AA29439@ukko.think.com> 
Date: Fri, 12 Apr 91 08:39:12 -0500
Status: RO

Guy,

   You've raised a very good question.  I'll let John Cowan or any of
the others who've worked with the formal grammar a bit more than I have
feel free to correct anything I've said here, but my understanding of
the situation is as follows:

1) The formal grammar with all of the terminators made non-elidible is
   LR(1) and hence unambiguous.

2) There is at least one case where eliding a terminator does result in
   ambiguity.  Unfortunately, the condition which determines if elision
   is permitted is context sensitive. For example "La Bab blanu" versus
   "Le fengu cu blanu".

3) In other cases (hopefully all of the rest) attempts to specify the
   elision in the grammar have resulted in complaints that the grammar
   is no longer LR(1).  Now, as you well know, non-LR(1) does not
   necessarily mean ambiguous.  I've done some investigation on the
   question of whether real ambiguities result in some of these cases,
   but I haven't reached any definite answers.

   All of this leaves us with a big research project, namely can we develop
a grammar acceptable to yacc (or some other parser generating system) which
will specify the language with elibible terminators.  Maybe I should have
made that my dissertation topic.

Dave Bowen