Why [h] does not get voiced

There is a minute detail of the phonology of Hungarian which
seems negligible  at  first glance,  but persistently annoys
phonologists concerned  with voicing assimilation: the ques-
tion of  why the segment [h]  (or rather  its non-prevocalic
allophone, [x])   fails to get   voiced  when followed by  a
voiced obstruent.   It  does not become voiced   despite the
fact that it  does itself trigger regressive voicing assimi-
lation, thus it participates  in the game typical of obstru-
ents only.  I will show that although the  target, [x] is an
obstruent,  it is not affected by  voicing because it is not
plainly voiceless like other obstruents, but aspirated.