Re: the machine grammars

[lojbab@grebyn.com (Logical Language Group)]

Ooops!  (I'm cc'ing this to everyone reveiwing, so I'll clarify).  Veijo
has caught one typo in the BNF, and asked a question that indicates something
I said in the text description of the changes is invalid.

1. There is an extraneous rule"sumti-tail<113>" in the BNF after
relative-clause<122>.  Delete it: it is a remnant of the older BNF.
The correct 113 rule is earlier, in proper numerical order.  Thus there is a
sumti-tail, a sumti-tail-1, and a sumti-tail-2 rule, in order, and should be
no other definitions of sumti-tail and its subcomponents.

2. Veijo asked about sumti-5<95>, which as written allows nested quantifiers
to the infinite degree on a sumti.  This is not a typo, since it appears
to be the saame in both the YACC and BNF grammars.  People may decide they wish
it to be a mistake{{,q and ask the reference rule to be changed to
"quantifier sumti-6", which allows only a single quantifier.

It appears that we did our usual and worked the grammar out to maximize
expressive capability.  In this case, the rule would allow a variation of
the multiply quantified indefinites that Colin didn't like.  Only in this
case, it is non-indefinites that get multiple quantifiers.  The grammar
would thus allow such things as "ci <vo le broda ku>" or "re <ci [vo le broda
ku]>".  Since relative clauses are not added until the next higher step,
at rule sumti-4, only the fully outer-most quantifier can have a relative
clause associated with it.

At first glance this would appear to be a useless wart, but I will let people
comment - if they see a use for it.  My suspicion is that it should be treated
as outside number selecting from inside number.  With Cowan not available thiss
week, I want to find out if he has any reasons supporting it.  I will not
fight strongly for it.