Message-Id: <MAILQUEUE2-102.931011110917.352@hylkn1.Helsinki.FI>
Message-Id: <199310110836.AA10092@kruuna.helsinki.fi>
Date:         Mon, 11 Oct 1993 09:35:53 BST
Reply-To: I.Alexander.bra0125@oasis.icl.co.uk
Sender: vilva
From: Iain Alexander <I.Alexander.bra0125@oasis.icl.co.uk>
Subject:      Re: TECH: Lean Lujvo and fat gismu
To: Veijo Vilva <VILVA@viikki21.helsinki.fi>
Content-Length: 4538
Lines: 89

This is a Big Subject, and I'm probably not going to do it justice
in this selmri:  How >do< you choose (the place structures for)
your predicates in a language based on the Predicate Calculus?

Certainly in Lojban, we have conflicting requirements.

On the one hand, we have a set of gismu which is intended to cover
the most commonly used predicates for expressing oneself in the language.
In this role, they want to have either places for _all_ the aspects
you might want to specify, or some subset of these which is judged
to cover the most common cases, and rely on the extension mechanisms
to cope with the rest.

On the other hand, these same gismu are also the _basis_ for these
very extension mechanisms, either by proxy (as it were), in tags using
the associated BAI, or in their own right in {fi'o}-constructed tags
or as components of lujvo.  In this role, it would probably make things
easier if they had a small number of places, typically two, which would
allow you to build up a more complicated concept from simpler pieces,
each new component adding one new sumti (which fills one of the component's
places), and relating it in its own way to the rest of the event
(which fills the other place of the two-place component relationship).

So, the price you pay for the (neat, elegant) economy of using the same
units for both purposes is the agony of juggling the place structures
to make it all work.  Tough.  There are no easy answers, at least not
in general.  We're going to have to make some compromises and allow some
leeway and hope for the best.

It's easy to construct horrendous examples for any particular choice
of place structures.  If I want to talk about the relationship
between a "text" and the language it's expressed in, I've first
got to decide what to use for "text".  Let's suppose I decide
I mean a {se cusku}, rather than anything more specific.
But {cusku} doesn't have a "language" place, so I've got to use
another word to tack this on.  We normally use {bau} or {bangu},
although this has a seemingly irrelevant-in-this-context place
for the people who speak the language.

(Various other gismu have a language place.  Most of them are
too specific - {[ve] tavla} is about a specific {ve cusku},
which isn't always appropriate, and {te/ve fanva} only applies
to a particular situation.  Most of the grammatical terms
(e.g. jufra, cmavo) have a language place, but are too limiting.
{se gerna}'s pretty close, but the one I really like is {te valsi}
- there are always some words involved.)

But suppose I decide to go with {bau/bangu} - it's traditional,
probably because it's the x1 that's interesting.  I've then got
to combine these two concepts, using something like
{[se] cusku bau la lojban.} or a lujvo {mi bausku dei la lojban.}.
This doesn't look like any of the standard lujvo-construction
patterns I remember, because the place structures don't connect
at all, and monjvo heuristics are going to give us a "people-
who-speak-the-language" place which isn't very interesting.
Perhaps {cusku} should have a language place, but then someone
would come up with a lujvo involving {cusku} where the language
was irrelevant.  It's a judgement call, and we're going to get
some of them wrong, and discuss them, and change them, and it's
a living language, isn't it?

I could do the same for the subject/topic of a statement.
Maybe {[se] notci} covers this, although it doesn't have
a language place.  {jufra} covers both, but is restricted to
a single "sentence" ({me la'e zo .i}?).  Etc...

It would no doubt be interesting to have a language which
allowed us to explore lots of different ways of constructing
predicates from (simpler?) parts, but I don't think Lojban
is it.  Lojban has several goals, and manages to cover most
of them pretty well.  It appears to have just about enough
mechanisms to allow constructions when they're needed,
but they're there as enablers, to allow us to get on with
the business of using the language, rather than as features
to be investigated in their own right.

I agree in principle with John(?) that gismu places should
all be there in the lujvo, but normally not be used.
What would be useful is some way of determining and/or
indicating which places are marginal - ideally they should
be displaced towards the end of the list.  But I don't see
any way of doing this.  Neither do I see how you {zi'o}
out places in a component of a lujvo.  (I don't see
{bangu zei be zei zi'o zei cusku} being practical, somehow.)

banzu
.i nunsipna tcika
mi'e .i,n.