home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.mail.mh
- Path: sparky!uunet!uaisun4!mrl
- From: mrl@uai.com (Mark R. Ludwig)
- Subject: Re: Question on Locking ...
- In-Reply-To: paul@engin.umich.edu's message of Thu, 10 Dec 92 18:41:11 EST
- Message-ID: <MRL.92Dec14063227@sun4.uai.com>
- Sender: mrl@uai.com (Mark R. Ludwig)
- Organization: Universal Analytics, Inc., Torrance (LA), CA
- References: <fAR=jkB@engin.umich.edu>
- Date: 14 Dec 92 06:32:27
- Lines: 51
-
- In article <fAR=jkB@engin.umich.edu> paul@engin.umich.edu (Paul Killey) writes:
-
- |> It seems like Sun's style of locking one's mailbox is to link a
- |> tempfile in /usr/spool/mail to /usr/spool/mail/user.lock.
-
- Yes, I believe this is so. It also uses one of the kernel's advisory
- locking facilities (lockf() or flock()), but I forget which. Despite
- delivering a non-working rpc.lockd all these years, I believe Sun
- still recommends using NFS for the mail spool, so I suspect the call
- they use is lockf(). If NFS really is involved, then all the better
- to hang your network at random, my dear.
-
- |> I presume this is to make things more robust when you NFS mount
- |> your /usr/spool/mail.
-
- I presume you are correct. I suppose if a fixed rpc.lockd ever
- actually makes it into all the vendors' deliveries, it might actually
- be reasonably robust to NFS-mount the mail spool.
-
- |> Would inc do this sort of lock if LOK_BELL were used, and lockldir were
- |> made /usr/spool/mail? If someone could say, it would save me some time
- |> and effort.
-
- I don't know what LOK_BELL is, but the mh-tailor(5) documentation I
- have (6.7.2) implies that "lockstyle: 1" will do this.
-
- |> What about just pitching the Sun Mail and /bin/mail, using Mail and
- |> /bin/mail done from source, and using flock()? Accessing
- |> /usr/spool/mail via NFS is not under consideration here.
-
- That's fine too. Just make sure that all the programs which can
- access the mail spool agree on the locking method. That's all that
- matters.
-
- |> There is also the consideration of getting popper, imapd, and the
- |> sundry other programs that update /usr/spool/mail all agreeing on
- |> locking.
-
- Yes. I'm in the process of making POP the only access method, and
- moving mail out of the spool to enforce it (actually just another
- directory under /usr/spool). This will make the "standard" mail
- utilities never work because they won't know where the mail is. I'm
- planning to use flock() for locking, because this is a Sun, and I
- don't have much confidence in lockf() until such time as Sun stops
- patching rpc.lockd. I'm planning to use Chip's "deliver" and the popd
- from the MH distribution; I wonder if I should look into popper? I
- someone would comment on this, I'd appreciate it.$$
- --
- INET: mrl@uai.com NIC: ML255 ICBM: USA; Lower Left Coast
- "A computer is one of life's joys; it follows simple rules. Just like
- children, adults need toys, only we like to call them 'tools.'" -- Dave Ross
-