Re: Dr. James Cooke Brown

["Bob LeChevalier (lojbab)" <lojbab@lojban.org>]

At 11:32 AM 02/19/2000 -0500, Robert A. McIvor wrote:
> >The posted YACC grammar does not include the lexer/preparser, which
> >contains (or hides) a substantial amount of grammar.  As of the last
> >version I had access to (and Trial 80's comments suggest that this is
> >true), I could write any random string of LWs (cmavo), precede it by a word
> >for a number and omit the space, and the entirety becomes a number
> >compound.  Likewise for a tense and probably some other compounds.  In
> >effect it means that there is no number or tense grammar since not all
> >spaces are lexemic pauses.  I understand that you RAM have tried to do some
> >work on this problem, but it seems to be a large problem with unknowable
> >side effects until the result is seen.  More than half of the Lojban
> >grammar is the YACC-encoded lexer grammar and the MEX grammar which in our
> >case is no longer primarily a lexer construct, and a large percentage of
> >our changes during the years before baselining involved the working out of
> >bugs in that grammar.
>
>         It is true that the lexer has not been  published.  While it is a 
> state
>grammar, it has not been put into a formal format and it was hand derived.
>Spaces are irrelevant to Loglan cmavo and, in fact, the lexer begins by
>removing all spaces between them.  You would find the same behaviour if 
>you had left
>the spaces in.

This is then a significant change from Scott Layson's parsers (circa T55?).

>   The reason a number is concatenated with many other little words like
>the PA (tense) lexeme, and the TAI (lettoral) lexeme as well as PO is that
>such compounds are intentionally grammatical in Loglan.  The grammar
>concatenates any collection of cmavo that constitute
>a single lexeme.

I understand - but in the comments as I read them (and in reality in the 
Layson parser), cmavo of ANY lexeme that were concatenated to PA and TAI 
formed a valid compound, so that "patailenehue" would have been taken as a 
grammatical Loglan compound of PA, but of course with the words written 
separately it would be ungrammatical randomness.  It sounds like this has 
changed, though the comments do not suggest it (pacenoina is considered PA2 
but includes the ce noi which is not PA - clearly there is a hidden grammar 
here, since pacenoina is valid, but pananoice should not be - the (T55?) 
Layson parser did not have a problem with either.

>   It is true that we do not yet have a full MEX grammar.

Again, it comes down to what you accept as a valid compound - if indeed the 
lexeme content of a MEX compound is limited that is an improvement over no 
content restrictions on compounds at all.

lojbab
----
lojbab                                             lojbab@lojban.org
Bob LeChevalier, President, The Logical Language Group, Inc.
2904 Beau Lane, Fairfax VA 22031-1303 USA                    703-385-0273
Artificial language Loglan/Lojban:  http://www.lojban.org (newly updated!)