On Sun, Jun 05, 2005 at 04:06:42PM -0700, Steve Lamb wrote:

>     It is not a silly response, it is factual.  500Mb of mail at an
>     average of 5Kb per message is 100,000 messages.  100,000 files in
>     a single directory is not "more efficient for individual deletes"

Okay, at the risk of starting a flame war, it's still silly. Allowing
users to have 100,000 messages in a single directory is insane, and is
purely the fault of the administrator for not forcing users to download,
sort, archive, or otherwise deal with their mail in a sensible way. Use
inode quotas, for goodness sakes. That said, a modern filesystem such as
XFS or ReiserFS gives pretty good performance even on directories with
tens of thousands of entries.

Moving 100,000 emails into a single file just shuffles the problem
around. Appending to an mbox is reasonably efficient. But instead of
inode consumption, you have CPU spikes when email applications try to
locate individual emails inside a single huge text file, or disk spikes
when someone deletes a single email out of the mbox and the other 99,999
emails need to get written back out to disk.

All that is somewhat irrelevent, though, in that maildir was not
designed to provide better performance (although it can, in some cases),
but to provide guarantees that all messages are safely written to disk.
It protects against file-locking issues, allows MUAs to flag messages
individually, and allows for manipulation (e.g. editing or deletion) of
individual messages without affecting others messages. In short, its
primary goal is data availability.

You've really made my point point for me: the benefit of mbox or maildir
is heavily dependent on its usage. In your case, because of the *way*
you and your users manage mail on your systems, mbox makes sense in your
environment. For other environments, maildir offers a different set of
optimizations that can improve performance and reliability.

Personally, I use both, for different reasons and in places where they
make sense. Always use the right tool for the job at hand; zealotry is
inefficient.

-- 
Re-Interpreting Historic Miracles with SED #141: %s/water/wine/g


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Reply via email to