En réponse à [log in to unmask]:

Ok, let me now make a (late) summary of Proto-Semitic phonology as I see it
(this is an attempt to put in one all the information given to me by Vasiliy
and Keith, and may not be correct, but at least it's an attempt :) ):


stops, affricates and laterals:
p  t  ts  tS  tl  k     ?
b  d  dz  dZ  dl  g
   t' ts' tS' tl' k'

   s          sl  x  H  h
                  G  3

w  r      j   l
m  n

I chose to put the maximal inventory here, so that if there's a big mistake I
would have to substract phonemes rather than add them. I also took the idea of
Vasiliy that /q/ was originally /k'/, because it saved me the need of adding a
voiced uvular stop (it's already such a big inventory :) ) and it nicely
preserves the symmetry. That's why I also kept distinct both the lateral
stops /tl/, /dl/ and /tl'/ and the lateral fricatives /sl/, /sl'/ and the
continuent /l/.


short vowels: a i u

long vowels: a: i: u:

diphtongues: aw aj

Now, for what this inventory would be for 1st century Arabic, here's my guess:
- /p/ already changed to /f/,
- /sl/, /sl'/ and /tl/, /tl'/ already merged (as /tl/ and /tl'/? that would be
nice to reinterprete clusters), probably /dl/ already changed to /l/,
- affricates still alive and well, as well as /k'/ (good for merging with Latin
clusters, especially because of the Latin tendency to palatalise clusters of
the form tiV).

This would give this inventory:

   t  ts  tS  tl  k     ?
b  d  dz  dZ      g
   t' ts' tS' tl' k'

f  s              x  H  h
                  G  3

w  r      j   l
m  n

a  i  u
a: i: u:
aw aj

Well, this is probably a mess, but it was worth the try. If you don't find to
much to complain about it, I'll begin my work with it :) . Personnally I like
it. Still, even if I don't think there will be much difficulty on adding
uvulars and pharyngeals to the Romance phonology, I still have difficulties
with the glottalized consonnants. Maybe some clusters of two stops simplified
(Greek borrowings would help a lot by adding not very Romance consonnant
clusters like /ps/ and /pt/ in front of words).

Well, what do you think of all this?