Andr� Balsa wrote:

On Tuesday 10 May 2005 23:13, Douglas E. Engert wrote:

With all the problems with the integration of Krb5, AFS, PAM,
and OpenSSH. I would like to bring forth *again* the concepts of
separating out the pam_krb5 from the pam_afs2 from the aklog.

The basic concepts are:

 o Use the vendor's pam_krb5 without any AFS code.

 o Provide a separate pam_afs that gets a PAG using syscall, or
   /proc and forks execs a separate program to get the AFS token
   passing KRB5CCNAME= from the pam_getenv to the program.
   The pam_afs2 has no AFS or Kerberos libs dependencies.

 o The separate program is your favorite aklog with whatever
   version of Kerberos and AFS you want to use.


Hello,

This is just a short comment on the above.

The idea sounds good to me. I wish we could have an open discussion of the above, without any prejudice in favor or against the proposed changes.

I also understand this is a suggestion for the direction of future developments. Who would be responsible for implementing these changes and maintaining the corresponding code is another matter, as I believe the present OpenAFS team already has a high enough workload.

I would suggest the OpenAFS needs to maintain the pam_afs2 code and the gafstoken routine. What has happened without this is some Linux vendors have developed pam modules for krb5, or krb5+afs, but not all vendors do this, thus leaving it up to the sysadmin.

OpenAFS already has a aklog, and I have the gssklog for systems
that don't have Kerberos exposed, and I have said it could be donated to
OpenAFS.

If OpenAFS can provide the kernel extensions, they certainly can provide
the simple PAM interface too.

There is not a lot of code here, two source files  pam_afs2.c has 324 lines
of code, and the gafstoken.c has 412 lines.




Thanks, regards,

--

 Douglas E. Engert  <[EMAIL PROTECTED]>
 Argonne National Laboratory
 9700 South Cass Avenue
 Argonne, Illinois  60439
 (630) 252-5444
_______________________________________________
OpenAFS-devel mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-devel

Reply via email to