IMAP debate/thread-that-wouldn't-die

[bscott at ntisys.com]

On Tue, 21 Oct 2003, at 8:43pm, ken at flyingtoasters.net wrote:
> Sure, a zillion directory entries -- with its corresponding huge number of
> inodes -- is an imperfect solution.

  Unless you're using a filesystem that doesn't use inodes.  ReiserFS comes
to mind.  Note that I haven't tried Cyrus+ReiserFS together yet, but
independent reports say it works very well.

> Note, however, that I agree with Ben that merely copying it into the same
> directory (and, presumably, the same device) would certainly cut down on a
> lot of trouble.

  Less then you think.  There's still a ton of I/O that needs to be done.  
The problem with the whole copy-to-/tmp dance is that many systems have a
relatively small /tmp partition, compared to the size of some user's
mailboxes.  /tmp is often on the root filesystem, or at least the main
system disk.  Meanwhile, mailboxes and user data are often on huge RAID
volumes.  So when you run a purge with UW-IMAP (actually, UW C-client), you
fill /tmp and Bad Things Happen.  By keeping the file on the same partition,
you at least avoid that problem.  But you still need to do all the I/O.

> Maybe there's a reason they decided against that: beats me.

  I suspect it's mainly just because that's the traditional use of /tmp.  
And, if /tmp is large enough, and /tmp is also on a separate physical disk
from the mailbox, it would be a performance win.

> But -- for a record third time -- I'll go along with Ben ...

  Just don't let it become a habit.  ;-)

-- 
Ben Scott <bscott at ntisys.com>
| The opinions expressed in this message are those of the author and do  |
| not represent the views or policy of any other person or organization. |
| All information is provided without warranty of any kind.              |