[5474] in Release_7.7_team

home help back first fref pref prev next nref lref last post

attach in 9.9

daemon@ATHENA.MIT.EDU (Greg Hudson)
Fri Jun 2 18:04:32 2006

Date: Fri, 2 Jun 2006 18:03:46 -0400
Message-Id: <200606022203.k52M3kGf014901@egyptian-gods.mit.edu>
From: Greg Hudson <ghudson@MIT.EDU>
To: release-team@MIT.EDU
X-Spam-Score: 1.217
X-Spam-Level: * (1.217)
X-Spam-Flag: NO

I've pulled athena/bin/attach, athena/bin/quota, and athena/lib/locker
into the 9.9 source tree and started hacking on them.  I've realized
I'm not quite sure what features we want to preserve and what we want
to get rid of.

Obviously attach won't be performing the actual mount, so the options
which control how or where the mount is performed must be desupported
(--explicit, --mountpoint, --setuid, etc.).  On the other side,
--printpath and -Padd definitely need to work.

That leaves a misty gray area containing:

  * Authentication: we could strip this functionality out and just
    rely on pam_krb5 to authenticate to the cells most lockers are in;
    users could still explicitly aklog to other cells.  Or we could
    leave it in and let people attach or add lockers to authenticate
    to them.  (That would also preserve the NFS authentication
    functionality, which could be considered a good thing or a bad
    thing.)

  * Zephyr subscriptions: attaching a locker could still subscribe to
    filsrv messages for that locker's server.  I don't know if ops
    still relies on the ability to send filsrv messages, or how
    important it is to them.

  * The attachtab file: explicitly attaching a locker could keep
    records in the system attachtab like it currently does, allowing
    us to preserve (sort of) the listing functionality of "attach"
    with no arguments or with the --host argument.  An attractive
    aspect of removing this functionality is that it allows us to
    clear the setuid bit.  One alternative to maintaining the system
    attachtab is adding a user-specific list of explicitly attached
    lockers somewhere.

  * zinit: The current functionality of zinit relies on both of the
    previous two features.  If attachtab functionality is removed but
    zephyr subscription functionality remains, zinit could get subs
    for just the homedir, but that would miss lockers attached in
    .environment, so it's not very satifsying.

Any opinions?

home help back first fref pref prev next nref lref last post