Hi,

At Tue, 13 Sep 2011 20:06:29 +0900,
Luis Lavena wrote in [ruby-core:39512]:
> Changing that right now for sure will break assumptions of lot of
> programs on Windows, more on newer versions of Windows where symlinks
> and hardlinks are more common (Vista, 7 and 2008 Server).

For a program which doesn't care of symlinks, nothing would
change.  It accesses a different directory, drive or UNC, even
if it considers the symlink would be local.  Symlinks are
transparent in that sense, unlike "shortcuts".  Hardlinks have
no way to tell which one is "real".

> > Other API results were varying across Windows versions, so
> > FindFirstFile was the last resort for stat.
> >
> 
> In the context of setting Windows XP SP1 (0x0501) as minimum version,
> perhaps this can be minimized?

Probably.  But I can't and don't want to check those variants all.

> > Since these come from the spec of Windows filesystem and
> > security requirements, so it's not acceptable to change the
> > behavior of File.expand_path, unfortunately.
> >
> 
> Wouldn't be responsibility of File.realpath to solve
> symlinks/hardlinks and shortnames?

It could be, if File.realpath had been introduced at that time.

-- 
Nobu Nakada