[OpenAFS] when openafs becomes a windows IFS

Rodney M. Dyer rmdyer@uncc.edu
Tue, 22 Jul 2008 21:04:23 -0400

At 04:37 PM 7/22/2008, Jeffrey Altman wrote:
>>To follow symlinks, or not to follow symlinks, that is the question.
>That is what question?

Just a joke in the characterization of Shakespeare.

>And to whom are you posing it?

No one in particular.

>A symlink is not an object that Windows knows how to describe.

I clearly understand the technical limitations of symlinks under 
Windows.  I love them.  No... I LOVE them.  Gods gift from heaven.

Strictly speaking however this is an aspect of POSIX vs. Windows FS 
capabilities.  Symlinks are application based pointers, and should not 
reveal themselves as directories to the underlying FS find/search API, but 
that is exactly how the original developers of AFS on Windows decided to 
use them.  The Windows 32 bit attribute on FS objects have no attribute bit 
that describes a symlink object (this was an extremely bad idea, or lack 
thereof, on Microsofts part).  Even if Microsoft allocated an attribute bit 
for one today, no application could immediately make use of it.  It is an 
unfortunately ugly situation, but one where Windows is better off just 
following them.

Following symlinks causes grief for those who don't remember that they have 
links to other volumes within their FS directory tree.  This is what I was 
trying to express.  I suppose it might be nice if a Windows file delete 
didn't work through symlinks.  I haven't really given it critical thought 
however as to how many things this would break.  Probably a lot.

I both love and hate symlinks.  Much more love than hate.  If only 
junctions worked as well.