Re: 'native' symlinks for 2K+

["Chris January" <chris at atomice dot net>]

> > > Is there any interest in an implementation of symlinks via reparse> >
> points for win2k and above?
> > Reparse points using the filter supplied with win 2k/xp aren't that
great
> > for doing posix-like symlinks. However, in my bookmarks I have some free
> > code for writing file system filters for Windows NT/2k/XP. Perhaps we
could
> > write our own filter? This would mean Cygwin symlinks worked the same
with
> > native programs as they do with Cygwin programs. (I will, of course, put
my
> > keyboard where my mouth is if there is interest in this).
>
> That's interesting.  Don't one need access to the IFS Kit which
> costs $$$?  If you have access to the IFS kit, there's the problem
> of the licensing.  Since it's not publically available, I guess that
> MS has some restrictive license which e. g. disallowes publishing
> of header file details given in the IFS kit.  IOW, if you write a
> filter using the IFS kit either nobody else can compile it or you'd
> have to release compatible header files and library stubs which
> would be against the license.
This is a GPL'ed version of the ntifs.h file:
http://www.acc.umu.se/~bosse/ntifs.html

> Hmm, I have a deja-vu here... as if I already asked that question
> about IFS moons ago...
>
> Another aspect is, would we have to install something (probably a DLL)
> into the Windows system folder to get an own filter or is it enough
> if the Cygwin DLL itself tells the system "hey, I'm responsible for
> reparse points of type foo" ?
It would need to go into the SYSTEM32/Drivers directory, IIRC.
>
> Anyway, we should really think twice about that.  We already have two
> methods of creating symlinks, one of them is accessible by native
> Windows (well, sort of) and both have the advantage to work on 9x and NT.
true.

Chris