En réponse à Benct Philip Jonsson :

>This is interesting as it is analogous to Japanese, where
>historically *p\ changed to /h/, but retained a [p\]
>allophone before *u (now /M/).  I suspect we have to do
>with an universal which prevents the delabialization
>of [p\] before a fully labialized vowel.  NB Japanese
>/h/ has a [C] allophone before /i/ as well!

I doubt it. In some dialects (Gasconic dialects IIRC), the f > h went all 
the way, including in front of [w] ("fire" is "huek" there).

Christophe Grandsire.

You need a straight mind to invent a twisted conlang.