svn commit: samba r24465 - in branches/SAMBA_3_2_0/source/libsmb: .

[Derrell Lipman]

On 8/15/07, Volker Lendecke <Volker.Lendecke at sernet.de> wrote:
>
> On Wed, Aug 15, 2007 at 03:59:26PM -0400, Derrell Lipman wrote:
> > Yes, this is always done.  The code this replaces already sorted DENY
> before
> > ALLOW and commented that such was required.  The documentation I found
> today
> > also says it's supposed to be like that.  Additionally, inherited ACEs
> are
> > supposed to come after non-inherited ones according to the docs.
> >
> > There's no mechanism in libsmbclient for stating that "I know what I'm
> doing
> > so don't sort these for me" but since most users won't know what the
> sort
> > order is supposed to be, we're better off sorting according to the docs.
> > Remember that this is emulating the xsetattr() interface.  It's not a
> native
> > interface to the SMB protocol.
>
> Well, I don't want to play devil's advocate here, but there
> is precedence for bad things happening (did anybody say
> NFSv4 ;-)) if people look at the Windows ACL documentation
> and not THOROUGHLY test what's going on. I'd be a lot
> happier if this sorting did not happen behind the scenes but
> within a separate service routine.
>

Removing the sorting would break backward compatibility.  We've always
sorted.  The sorting is now a bit more strict than it was, but we've always
sorted.  This means that users who have never cared about specifying the
correct order have had things work properly most of the time.  Removing the
sorting would suddenly break their applications.

I don't consider it acceptable to break backward compatibility, so I don't
want to remove the sorting.  I will, however, look into whether there can be
some option for specifying that no sorting should occur so that people who
really know what they're doing have the option of specifying such and
getting around the automatic sorting.

Cheers,

Derrell