Re: TECH: more on morphology problem - some opionions

[John Cowan <cowan@SNARK.THYRSUS.COM>]

la lojbab. cusku di'e

> 2. A new phonological category of "permissible initial cluster" be defined,
> to apply to le'avla.  Clusters at the beginning of words would be required
> to  be such that all pairs of consonants within the cluster be permissible
> initial pairs, per the standard set already defined.

My preliminary analysis suggests that this proposal is safe.  The standard
flavors of type 3 le'avla include the following (where R = r,l,n):

        CVCCRC...       CCVCRC...       CVCRC...
        CCVRC...        CVVRC...        CV'VRC...

The first two employ 4-letter rafsi; the remaining four use 3-letter rafsi.
The forms CVVRC... and CV'VRC... may not be safe in all circumstances:
if the C... tail (the borrowed part) looks like a lujvo, the whole word
is invalid as a le'avla but valid as a lujvo.

The cases beginning CC are safe a priori; nothing can fall off.  The other
cases would, if CV were to drop off, begin RC, CRC, or CCRC, none of which is
valid by the rule above, because RC is never a valid initial pair.

However, I would propose an even simpler, safer rule which can't blow up
on us later, and somehow feels "more Lojbanic" to me.  Don't allow consonant
triples or larger multiples at the beginning of words at all.  They aren't
allowed anywhere else in the language (barring names, which have weird
phonotactics anyway).

> 3. A self-consistent definition of permissible medial cluster would be
> devised that covered these messy combiantions that can arise in le'avla.
> the current definition in the Synopsis was written only with gismu and
> lujvo in mind, and hence is vague and/or forbidding of the norm for lujvo.

I don't understand "vague and/or forbidding of the norm".  As I understand
it, the current definition of a permissible medial triple is that it must
take the form C1C2C3, where C1C2 is a permissible medial pair and C2C3 is
a permissible initial pair.

> 4. Cowan will investagate, and may propose some norms for final clusters,
> which would set a non-mandatory standard for use in names.

This is a wholly separate issue, and one which needn't be solved now.
Names are still just ...C<pause>, with no impermissible medial consonant
pairs.

--
John Cowan      cowan@snark.thyrsus.com         ...!uunet!lock60!snark!cowan
                        e'osai ko sarji la lojban.