[1298] in SIPB-AFS-requests
Re: s:a on too many acls
daemon@ATHENA.MIT.EDU (mhpower@MIT.EDU)
Mon Apr 4 00:36:25 1994
From: mhpower@MIT.EDU
Date: Mon, 4 Apr 1994 00:34:04 -0400
To: tlyu@MIT.EDU
Cc: sipb-afsreq@MIT.EDU
In-Reply-To: "[1296] in SIPB-AFS-requests"
> ... I think the correct course of action is to remove s:a from
>top level acls in user homedirs and notify all users involved about
>this. Comments?
I disagree. I think that once a volume has been created and handed
over to a user, it's not appropriate for s:a-types to make *any*
changes to the acls, even if it's believed the acls might have
originally been created by someone on s:a as a mistake. Users might be
relying on the acls for some unusual reason, and might simply not want
the 'Last Update' time on their volume changed by someone else...
Finally, I should mention that many of the user volumes that have s:a
access are the ones created in the original migration of homedirs from
charon to AFS by rfrench in August, 1989. This was discussed in
transactions [0032] - [0034]. A lot of these directories are sort of
unnecessary, in that the people have been gone for years. That is, the
acl problem might be better solved with 'vos remove' than 'fs sa' :-)
Using vos examine, you can generally tell what directories haven't
been touched in years. One obscure fact: volumes last updated around
5-6am on May 24, 1992 were the result of ckclark chowning the volume
mountpoints (see [0691]) -- the last user access was often a few years
before that, even. This points out yet another reason not to modify
user volumes ... we lose the last-update data and can't easily tell
whether a homedir is "abandoned" or not. If it's really necessary to
change users' acls, please at least save the last-update date in a
file somewhere, in case we need that information later...
Matt
