XFS acl support for SAMBA_2_2

[Jeremy Allison]

On Mon, Feb 19, 2001 at 03:44:30PM -0500, John Trostel wrote:

> I'm doing this using the Linux/XFS CVS from SGI.  It's a fairly straight port
> of the SGI XFS on IRIX to Linux and therefore doesn't get involved in a lot of
> the thing that Andreas is working on.

Yeah, but it *should*. See below.

> I've read over the new draft and am not sure how this will (or should) affect
> the implementation of Linux XFS ACLs.  Linux XFS acls are implemented with
> system calls:
> 
> int acl_get (const char *path, int fd, struct acl *acl,
>                     struct acl *dacl);
> 
> and
> 
> int acl_set (const char *path, int fd, struct acl *acl,
>                     struct acl *dacl);
> 
> Userspace functions currently defined are very POSIX compliant and include:
> .... snipped.....
> These are very similar to the SAMBA/POSIX defaults and the Solaris/HPUX/etc.
> acl functions already in use.

Yes, but the big problem is we now have *two* implementations
of *almost* the same API in userspace, and two different 
implementations in the kernel, depending on which filesystem
you're using.

This is *BAD* - and only one will get added to the mainline
kernel, I'm sure of that (or I don't know Linus & Ted very
well :-) :-). In addition, which userspace API gets added
to glibc ? This is a receipe for disaster.

SGI/XFS and Andraes need to work together to ensure that the
POSIX API (which Andraes's work is exactly created from) is
the one that XFS uses to get/set ACLs.

I'm hoping that the ACL in Linux efforts get merged as soon
as possible - or we're re-inventing vendor forking on *one*
system.... :-(.

Jeremy.
-- 
--------------------------------------------------------
Buying an operating system without source is like buying
a self-assembly Space Shuttle with no instructions.
--------------------------------------------------------