[OpenAFS] Many questions

Hartmut Reuter Hartmut.Reuter@rzg.mpg.de
Thu, 14 Dec 2000 11:49:58 +0100

"Patrick J. LoPresti" wrote:
> (If any of these are answered elsewhere, please feel free to flame me
> and give a pointer.)
>   What is the real status of Open AFS as far as IBM is concerned?  (I
>   have read the FAQ, but I am curious about the reality.)  Will IBM be
>   working to make fixes and improvements flow from IBM AFS to Open AFS
>   and vice-versa, or is this a one-time code fork into two trees which
>   will diverge from now on?

The OpenAFS advisary board has promized not to break compatibility to

>   IBM says that Open AFS is "unsupported", which is reasonable.  But
>   what if, hypothetically, I wanted to license Windows clients but use
>   OpenAFS servers?  Would I still be able to get support for the
>   clients?

You may build your Windows client in OpenAFS as well.

>   The Quick Beginnings guide makes no mention of a special "fsck"
>   program for Linux.  Does this mean that the standard e2fsck is
>   appropriate for checking Linux /viceXX partitions?  (It sounds like
>   it, but I want to make sure.)

The AFS-server in Linux uses the so called NAMEI-Interface: i.e. all AFS
files and directories are regular files in the filesystem which,
however, have synthetic names. Therefor no special fsck is required.

>   Are the Linux OpenAFS modules compiled with symbol versioning?
>   On a related note, is there any effort underway to get the Open AFS
>   kernel modules integrated into the mainline Linux kernel sources?
>   Would the license on Open AFS even allow it?

There were already some discussions about this point in the mailing
list. Since the AFS kernel extensions are huge compared with the Linux
kernel itself and since the kernel versions are develloping rather fast
it seams for the moment better to keep things separate. Also, the Linux
port should not fork from the general AFS client code.

>   The RPMs do not seem to include modified versions of the rsh suite
>   (or any similar tools).  The source for rsh.c (OpenAFS 1.0.1
>   distribution) defines a local "pass_tokens" variable but never uses
>   it for anything; in other words, the source code itself seems to
>   make no sense.  I understand that I can use PAM to allow various
>   services to authenticate against the AFS Authentication Server, but
>   what I want is to be able to pass authentication tokens *without*
>   typing my password again...  Is that possible, even in principle?

This is old stuff. You better use ssh instead of rsh. We have local
changes to ssh which do a token transfer.

-Hartmut Reuter

> Thanks in advance for any answers.
>  - Pat
> _______________________________________________
> OpenAFS-info mailing list
> OpenAFS-info@openafs.org
> https://lists.openafs.org/mailman/listinfo.cgi/openafs-info

Hartmut Reuter                           e-mail reuter@rzg.mpg.de
					   phone +49-89-3299-1328
RZG (Rechenzentrum Garching)               fax   +49-89-3299-1301 
Computing Center of the Max-Planck-Gesellschaft (MPG) and the
Institut fuer Plasmaphysik (IPP)