Hi Greg,

  You wrote:

>One issue I have had with accessing coda over samba is that samba
>seems to open files for read/write, even if the underlying operation
>on Windows should have been read only (i.e., if Windows is sensible,
>which I realize is a counter-factual hypothesis).  This results in
>...
>IMHO this is a bug in coda, but I realize it is hard to fix.
>(Meaning, "open for read/write, read, read, close" should not cause a
>store.)

Actually, I consider this a bug in most of the Windows software.  This
is something that happens all the time.  An application opens for
r/w and then only reads.  And the application knows it will open for
read only.  I put in a solution for the NT/2K/XP kernel module in that
it only passes an "open for write" to venus when an actual write happens.

Now, it is true that Coda could easily put in an optimization for
this case.  The kernel modules could call venus with a flag that
says if any actual writes were done.  I suspect it would not do much
other than change the kernel-venus upcall for close.  In fact,
we could add a "no-write-close" upcall if we didn't want to break
the current close upcall.

--Phil

-- 
Phil Nelson                       NetBSD: http://www.netbsd.org
e-mail: phil@cs.wwu.edu           Coda: http://www.coda.cs.cmu.edu
http://cs.wwu.edu/faculty/nelson