[OpenAFS-win32-devel] @sys variable name

Douglas E. Engert deengert@anl.gov
Mon, 05 Apr 2004 10:18:44 -0500


Jeffrey Altman wrote:
> 
> Tim C. wrote:
> 
> >>In Feb 2004, IBM dropped support for Windows NT4 systems.
> >>As such they renamed the sysname from "i386_nt40" to "i386_win2k" as
> >>documented at:
> >>
> >>http://www-306.ibm.com/software/stormgmt/afs/manuals/Library/patch.readme.htm#HDRWINDOWS_PRODUCT_NOTES
> >>
> >>What is the opinion of the community regarding this change?
> >>
> >>
> >  I was actually always suprised at the fact that windows didn't do
> >this.  I would be all for this.  People that use this regularly, but
> >don't want to distinguish between them, can just place links from
> >i386_win2k to i386_nt40.  We've done this for a number of other OSes.
> >One question I would through out there is, why not i386_nt50?  I mean
> >sun is still sun4x_59(not sure what the sysname for 10 is), in spite of
> >the OS being solaris 9.
> >
> >
> The value of sysname is determined at compile time or is loaded
> from the Registry at HKLM\SOFTWARE\...\TransarcAFSDaemon\Parameters,
> SysName.
> 
> I am not a long time user of AFS so I do not really understand if
> this value is supposed to represent the version of the AFS Client
> Software or the version of the operating system it is running on.

Sort of what version of the OS kernel the cache manager was compiled for. 
But there have been times where the OS may have been updated, but the 
older AFS release would still run on the new kernel so no change was 
needed in the @sys. The fs command can also be used to set the sysname
on a client. 

Many sites would then use this for different builds of applications 
on UNIX at least, as the @sys would change with each release. 
Very handy for build directories for applications with common source
or common scripts to be used across platforms.    

> If it is supposed to be the version of the operating system then
> I would suggest that we not use a compile time option but instead
> provide for a value constructed at runtime based on the OS version.
> We would probably want to be able to distinguish:
> 
>     * architecture - 32x86, 64x86, itanium, ???
>     * operating system - nt40, w2k, xp, w2k3
>     * os variation - home, pro, server
> 
> Or are these variations really not useful to people?

That might be overkill, but could be usefull.   

 
_________________________________________
> OpenAFS-Win32-devel mailing list
> OpenAFS-Win32-devel@openafs.org
> http://lists.openafs.org/mailman/listinfo/openafs-win32-devel

-- 

 Douglas E. Engert  <DEEngert@anl.gov>
 Argonne National Laboratory
 9700 South Cass Avenue
 Argonne, Illinois  60439 
 (630) 252-5444