I have not read Chalmers' book, but I have read several of the papers on
his web site and I've seen discussion on comp.ai.philosophy.
I agree with you that Chalmers' notion that a zombie world is "logically"
possible is questionable. He eventually rejects the possibility, as you
say. But he is willing to entertain it as a hypothetical starting point.
The question of whether zombies are logically impossible versus merely
physically impossible is a bit too subtle for my grasp of philosophy.
It seems to me that the kinds of arguments philosophers make are basically
logical and don't really depend on the nature of the physical world.
I suppose they are not logical in the strict sense of axiomatic deduction;
even in the fading qualia example Chalmers admits that it is conceivable
that consciousness would behave in that bizarre and counter-intuitive way,
where people are unable to describe their fading consciousness.
Is it this loophole in his argument that forces him to leave open the
possibility of zombies? In that case I would say that the problem
is that his argument is not airtight, rather than that it is based on
the nature of the physical world. The conclusion I draw is that it is
*very likely* that zombies are logically impossible, not that they are
logically possible but physically impossible.
Hal