Digit strings

["Mark E. Shoulson" <cbmvax!uunet!CTR.COLUMBIA.EDU!cuvmb.cc.columbia.edu!shoulson>]

Chris Handley and I have been having an off-line discussion about the
digit-string nature of Lojban numbers, and how it's hard to tell how big a
number is at the beginning, and easy to lose track of digits.

In English, when speaking a number, we periodically remind the listener of
the approximate logarithm (though in a somewhat erratic fashion), saying
something like "one million, two hundred thirty-four thousand, five hundred
sixty-seven".  As a listener, I might not remember all the digits spoken,
nor even how many there were, but right at the outset I heard "one
million", and that tells me what I find most important about the number:
its most significant digit and its charactaristic.  Even "one hundred
twelve million..." isn't so long that I get lost before hearing the
decimal-point fixing "million".  But when reading a number, we're in more
trouble.  For small numbers, we have no problem deducing the rough size at
a glance, but even with commas, when presented with a number like
143,238,284,823,183,097,912,836,284,284 I'd be reduced to counting out
groups of three ("hundreds, thousands, millions, billions (I'm American,
remember?), trillions....") before I came to the estimate, "Oh, it's
something over one hundred forty-three octillion" (using American
reckoning).

Using the numerical comma, {ki'o}, Lojban can sometimes help shorten
numbers that end in a mess of zeroes, yielding "paki'oki'o" for "one
million" and so on.  But since {ki'o}, unless followed by nothing or
another {ki'o}, is just a number comma ({paki'omucire}=1,532), it's no more
helpful that the commas in the huge number above.  Even worse, because
it'll be that way even when spoken, when I can't hunch over the screen/page
and count 'em.  And what about {paki'oki'orecivoki'omuxaze}="1,,234,567"?
Is this "1000234567" or "1000000234567" or "1234567"?  I'd favor the
last, as the other two seem of limited utility and this at least gives us a
way to indicate the size of the coming number a *little* more succintly as
we go along.  Chris's wish that we have cmavo related to words like
"thousand" and "million" and so on would be nice, except I don't think we
can spare them (OK, granted that {ki'o} is related {kilto}=1000), and
besides, where would we stop?  I briefly considered something about
subscripting the commas, but the grammar wisely doesn't allow that.
Subscripts in a digit string would be scary.

What do you all think?  (ju'i rodo do'u pe'ipei)

~mark