Thanks again Martin for the guidance.  It proved quite helpful in completing 
this patch.

This latest version (3) compiles and has been tested under Windows Vista and 
Windows XP and behaves as expected (with a NotImplemented error under XP).

I did find that I had to tweak the behavior of stat.S_ISLNK because my 
implementation of the st_mode bits for a symlink has different values than 
that of a posix call.  I don't fully understand the implications of my 
implementation.

An alternate implementation would be to have set_symlink_stat[AW] reset all of 
the st_mode flags except for 0o120000, to be consistent with what is returned 
by a posix stat call.  This approach doesn't seem quite right either.  I would 
appreciate a comment on this by someone more familiar with this type of 
problem.

Other than that issue, what else needs to be addressed to get this patch 
integrated into the 3.1 or 3.1.1 release?