Message-Id: <9110112049.AA28932@relay1.UU.NET>
Date: Fri Oct 11 20:21:14 1991
Reply-To: cbmvax!uunet!pucc.princeton.edu!jimc
Sender: Lojban list <cbmvax!uunet!pucc.princeton.edu!LOJBAN>
From: cbmvax!uunet!pucc.princeton.edu!jimc
Subject:      BAI places
To: John Cowan <cowan@snark.thyrsus.com>, Ken Taylor <taylor@gca.com>,
        List Reader <ghsvax!hal>
Status: RO

(In an attempt to reduce list pollution I replied directly to Mark
Shoulson's response to my posting on this topic, but it looks like the
item may become of more general interest.)

I asked if all possible BAI/PU/FI'O places were considered to be part
of the relation of every selbri.  John Cowan says yes.  Mark Shoulson
says that it's ridiculous for a simple motion word bridi to have a
place for the language in which it is expressed, the name which it (the
bridi) has, the thing it is the name of, etc. etc., not to mention the
FI'O's.

Nonetheless, it's official that you can set a tense default and have the
effect that the tense is dumped on every bridi (or just jufra?).  I also
feel (this isn't official) that the listener and speaker are equally
important and equally present on every jufra whether or not specified
by words.  So some BAI places are universal; these three are probably
not the only ones.

Lojbab argues strongly, and rightly in my opinion, that we should not
designate a specific set of _cases_ that are different from the rest;
we can't justify picking particular places and not others.  So all BAI
places have to be included.  But under Mark Shoulson's blandishments
I find my support wavering.  Particularly appalling is the idea of
predicates with infinitely many places!

Here's a compromise: each predicate has a list of essential places, and
every instance has all these places whether or not specified by words.
But only some of them are numbered; the rest are served by BAI/PU/KI'O.
For example in klama, I would number the standard motion word arguments
of mover - destination - origin - route, and let the "means" be served
by a BAI phrase with sepi'o.  Unlisted BAI's if present would be
interpreted similar to a restrictive subordinate clause on the bridi,
not as an actual part of the relation.  For example, the "speaker"
case, even though applicable to every bridi/jufra, is really not
properly part of a typical bridi, but expresses a separate relation
between the jufra and the person who said it.

By pruning the places we accomplish several goals:
    1.  We avoid severe philosophical and doctrinal problems with
        infinite numbers of places and bizarrely irrelevant places.
    2.  Regularities can be made more clear both in the numbered places
        and in the essential BAI's.  This aids learning.
    3.  By specifying a transformation of the unlisted BAI phrase to a
        predicate form, we make it more nearly possible to analyse
        Lojban entirely via predicates, which is the proper way for a
        predicate language.  But we avoid the deleterious effect such a
        transformation would have on essential BAI's.
    4.  The fewer "extra" numbered places, and the more regular the
        remaining places are, the easier it is to build dikyjvo.

                -- jimc