--- On Thu, 1/20/11, BPJ <[log in to unmask]> wrote:

> Some further observations: 

> 1)  You may of course have the same types of changes
> in
>     final/preconsonantal *w as well, so that you
> get *-ew > -ej,
>     *-ow > -o/-u, *aw > -a/-o.
> 2)  If you are concerned about getting a lot of
> diphthongs and VjV
>     you can always have various
> monophthongizations later. Those are a
>     darn good source for dialects too. And
> *Vju/ujV/iw/ew are possible
>     sources for /y/s.
> 3)  If you have changes like *ew/*ow/*aw > /o/ but
> end up with
>     very many /o/s then maybe the language had an
> /a/ ~ /Q/ ~ /o/
>     distinction which the script failed to
> express, conflating /Q/ with
>     either {a} or {o}.
> 4)  Another possible change is /w/ > /g/ or /w/
> > /b/, although
>     neither of these strikes me as likely as an
> across-the-board
>     change: *wa > /ga/ seems the most likely,
> with decreasing
>     likelihood before front vowels, *o and *u.
> Initial *wo> /go/,
>     and especially *wu > /gu/ seems to me less
> likely than the same
>     intervocalically, and final/preconsonantal *w
> > /g/ seems the
>     least likely. Also *wo, *wu > /bo bu/ seem
> more believalble than
>     /go gu/.

(snip some)
A very nice discussion of the possible fates of *w. Proto Austronesian *w takes part in many of these.
==loss in many languages, either outright or just in certain position (e.g. intervocalic, as in Malay, Indonesian)
== > h / #__ in Malay, Indonesian; some South Sulawesi lgs. show /h/ in all positions, either as outright or dialectal change
== > (written) b  or v (probably for [B]
== > gw in Chamorro (Guam) (I suspect early Spanish orthographic influence-- it may just have a noticeably velarized quality, as Engl. does and apparently proto-Germanic *w did (cf. the development of Gmc. loans in Romance). 

In my favorite sub-family (Lettic/Luangic), it is variously preserved as /w/ in some or lost outright; lost in certain environment e.g. *i_a; in the Wetan lg. it > IPA [j], in a lang/dialect called Moa, it is (written) g, always a puzzle to me that i ascribed to some 19th C. Dutch pronunciation, but which the Wetan investigator (a highly competent linguist) described as "similar to [j] but with more friction", from which I assumed it must resemble Spanish /y/, CSX [j\] IIRC. 

And oddly, in the possible final *-Vw combinations, it's lost in *aw and *iw; contrast *-ay > e but *-uy > i. But these behave oddly in many AN languages.