I just saw this note on the reiserfs mailing list. Is there any truth to it?
i.e. should I not expect to see Solaris 8, Win2k, etc. releases of AFS?
-- Nathan
------------------------------------------------------------
Nathan Neulinger EMail: [EMAIL PROTECTED]
University of Missouri - Rolla Phone: (573) 341-4841
Computing Services Fax: (573) 341-4216
-----Original Message-----
From: Michael Gerdts [mailto:[EMAIL PROTECTED]]
Sent: Friday, June 16, 2000 10:43 AM
To: Xuan Baldauf
Cc: Albert D. Cahalan; Christopher Browne; [EMAIL PROTECTED]
Subject: (reiserfs) Re: Alternatives to NFS
On Fri, Jun 16, 2000 at 03:39:53PM +0200, Xuan Baldauf wrote:
> > You forgot Arla, the AFS clone.
>
> As the Arla site stated, currently there is no free and working Arla
server
> available. Maybe this changes.
Since IBM bought Transarc, all development of AFS has been moved from
Pittsburgh, PA (US) to India. All of the original Transarc people have
left (expcept for the person that is training the new people in India),
leaving the only working AFS server implementation in the hands of people
that don't know the code. Furthermore IBM considers AFS to be in
maintenance-only mode. Expect no enhancements.
I really hope that some of the Linux-friendly people in IBM start putting
pressure on the AFS and DCE/DFS folks to open the source to these
technologies. This would surely boost the feature set of AFS in short
order. (e.g. Most sites seem to have Kerberos V patches that they apply to
their copy of the purchased AFS source code. AFS only has Kerberos IV.)
But... this has very little to do with reiserfs. The AFS file server, as
it is currently implemented, does not store the files on the server in any
sort of a usable format. There is only one 76 byte file that serves as a
handle for each volume. Thus:
% ls -l
total 806
-rw------- 1 root root 76 Jan 5 09:56 V0536870913.vol
-rw------- 1 root root 76 Jan 5 10:10 V0536870916.vol
...
-rw------- 1 root root 76 May 4 11:47 V0536873537.vol
-rw------- 1 root root 76 May 17 04:08 V0536873539.vol
-rw------- 1 root root 76 Jun 7 12:22 V0536873585.vol
drwx------ 2 root root 8192 Dec 29 11:04 lost+found/
% du -sk .
405 .
% find . -print | wc -l
find: cannot read dir ./lost+found: Permission denied
396
% df -k .
Filesystem kbytes used avail capacity Mounted on
/dev/vx/dsk/rootdg/volafs2ca
24860042 18649197 3724841 84% /vicepca
The volumes contained on that file system actually contain 951395 directory
entries, some of which are effectively just hard links between RW, RO, and
backup volumes.
Additionally, AFS uses directory ACL's, not file ACL's. As such, hard
links between directories fail (oh no, not that discussion again!) and it
is really difficult to come up with a way to restrict access to files that
need to have different permissions yet are in the same directory. That is,
/etc/shadow and /etc/motd under an AFS-style ACL scheme would have the same
permissions. (Not exactly true, but very close.)
AFS has some nice features such as client-side caching, strong
authentication (kerb IV), and replication between servers. It does have
plenty of weaknesses as well and should not be assumed to be wholely better
than NFS, SMB, or NCP.
Mike
