Re: Announce: MaildirSync

[dLux <dlux@xxxxxxx>]

<quote from="John Goerzen" on="Sat, Oct 12, 2002 at 02:38:31PM -0500" mark="|">
  | 
  | Can you spell out exactly how this differs from OfflineIMAP?  The
  | description sounds similar to me :-)
  | 
</quote>

The main differences:

- It synchronizes Maildir with Maildir, not IMAP with Maildir.
- It is written in perl, not python.
- Maildirsync uses a two-process model (source + target) like rsync, so
  perl and maildirsync must be installed on both sides.
- It uses two different state file for both synchronization state.
- Currently it does not have so much features, like configuration file
  reading, message filtering, etc.
- It uses very small amount of data transfer.
- The protocol is a "batch-like" protocol: one side sends a lot of data
  and the other side sends a lot of data and then they commit changes.
  This is better than any interactive protocol, because we can send
  larger packets which can be compressed better, and we don't have to
  wait for the answer for every command like the interactive protocols
  (like IMAP or POP3).
- MD5 sum of the message _body_ can further reduce the required
  bandwidth if you move messages between folders.
- VERY fail-safe. It assumes that the maildir folder change any time, so
  it does not crash or does anything bad if it does not find a message
  in a place where it was some time ago. It simply ignores it and next
  time it will propagate the changes to the other side.

These are the main differences.

I have a mailbox with more than 9800 email and the synchronization in
any low-bandwidth network takes about 25-30 seconds if there are only
small changes.

If you run it first time, it will calculate the MD5 sum of each message
bodies, so it is quite slow (it tooks 5-5 minutes on my comuters for
both sides).

Szabó, Balázs (dLux)
--
 "I did a 'zcat /vmlinuz > /dev/audio' and I think I've heard God..."
                                          (mikecd on #Linux)