=encoding latin1

=head1 NAME

maildirsync - Online synchronizer for Maildir-format mailboxes

=head1 SYNOPSIS

    maildirsync.pl [ --recursive ] [ --backup path ] [ --backup-tree ]     \
        [ --bzip2=bzip2 ] [ --gzip=gzip ] [ --maildirsync=maildirsync ] \
        [ --rsh=ssh ] [ --verbose ] [ --alg=md5 ] [ --delete-before ]      \
        [ --exclude=^/Folder1 ] [ --exclude=^/Fold.*er2 ]                  \
        [ --exclude-source=^/Folder3 ] [ --exclude-target=^/Folder4 ]      \
        [ --rename="s/SourceFolder/TargetFolder/" ]                        \
        [ --destination=win|lin ]                                          \
        [ -r ] [ -b path ] [ -B ] [ -R ssh ] [ -v ] [ -a ] [ -d ]          \
        source dest state_file.bz2

A simple two-way synchronization:

    maildirsync.pl -rvv -a md5 desktop:Maildir Maildir lib/sync_desk_note.bz2
    maildirsync.pl -rvv -a md5 Maildir desktop:Maildir lib/sync_note_desk.bz2

=head1 DESCRIPTION

maildirsync is a utility for online Maildir-synchronization. It is designed
to be used on live maildir folders, be fail-safe and optimized for minimal
bandwidth.

=head1 HOW IT WORKS

If you call the program once, it will propagate the changes from the source
side to the target side. Two-way synchronization requires two state-file
and two call for the program.

This propagation is basically two different operations from the source side:

=over 4

=item *

New file is created or an existing file is moved to a new location (e.g flags
are changed).

=item *

A file is deleted in the source side. (Will be deleted in the target side
also).

=back

At the first phase, the source side reads the state file (which stores the
state of the last synchronization) and compares it to the current state.
It collects the changes and sends them the target side.

The target side checks every file, which is marked new in the source file,
and decides if:

=over 4

=item *

The file needs to be downloaded fully

=item *

Its header needs to be downloaded

=item *

We have this file already, so we need no data.

=back

After it decides, it sends back the requests for new files. 

Then the source side will send the files to the target side, which stores
them into the Maildir structure.

After the send operation is completed, both operation agrees upon the
commit operation. Then the source side saves the new state into the state
file.

Note if we forget saving the state, or the program exits before it, the
operation can be restarted without data loss and inconsistency, because all
operations can be redone without errors.

=head1 REMOTE OPERATION

Maildir can be used in remote mode, so it can synchronize Maildir folders
between computers. If you want to use it this way, you have to provide the
host name before either the source or the target, like:

    maildirsync.pl ... desktop:Maildir Maildir lib/maildirsync.bz2

or

    maildirsync.pl ... Maildir desktop:Maildir lib/maildirsync.bz2

In remote mode, the target side must have maildirsync installed also. (See
the --maildirsync command-line argument).

The state-file must be in the same system as the source, so the source file
in the first example is searched in the "desktop" computer, and in the
local computer in the second example.

At least source or the destination must be local, so you cannot sync
maildirs on two different remote hosts.

=head1 COMMAND-LINE SWITCHES

Some command line switches has two form: a short form and a longer form. In
the short form, the switches can be grouped, like: -vvvr. Short options
with parameters also can be grouped, but the parameter must be the
following command-line argument, like:

    maildirsync.pl -rvvvbR Maildir/Trash/cur ssh ...

It is the same as:

    maildirsync.pl --recursive --verbose --verbose --verbose --backup \
        Maildir/Trash/cur --rsh ssh ...

Long options can use '=' for assigning the parameter, or they can use the
syntax above.

Let's see what switches we have:

=over 4

=item --recursive, -r

Process the base folder as the recursive collection of Maildir folders.

=item --backup dir, -b dir

The deleted files are backed up to the specified directory (this directory
does not needs to be a maildir folder). The directory is relative to the
start-up directory, not the target base folder! The directory is created if
it does not exists.
  
=item --backup-tree, -B

This option is only useful when used in conjunction with the --backup option.
If set, deleted messages are moved to a Maildir folders tree inside the backup
directory with the same relative path. The resulting backup directory can be
used with any Maildir-capable application (MUAs, MTAs, etc.).

=item --bzip2 bzip2

Path to the bzip2 utility (used only when the state file has .bz2
extension). Note that using bzip2 is turned out to be quite unstable, it
sometimes leave the state file empty or corrupted.

=item --gzip gzip

Path to the gzip utility (used only when the state file has .gz extension).

=item --maildirsync maildirsync.pl

Path to the maildirsync utility on the remote machine (if we use maildirsync
in remote mode).

=item --rsh ssh

Path to the utility, which can be used to connect to the remote side. It
defaults to "ssh". Note that the protocol, which is used in remote mode,
does not contain compression, but the data can be compressed well, so I
suggest using the ssh compression for this purpose.

=item --destination

If it is provided, then it does transformation in the file names. The
transformations are the following:

=over 4

=item *

win: Converts ":" to ":" (since windows don't allow colon in the filename).

=item *

lin: Converts ":" to ":".

=back

=item --verbose, -v

Adds more verbosity to the operation. There are currently 6 different
verbosity level:

=over 4

=item C<0>

No information at all.

=item C<1>

Main operations.

=item C<2>

Files sent, received, deleted, moved, md5 calculations.

=item C<3>

Directories read and created.

=item C<4>

Options + command echo.

=item C<5>

Misc info about file transfer.

=back

=item --alg md5, -a md5

Selects the synchronization algorithm. Currently two algorithms are provided:

=over 4

=item id (default)

The messages are synchronized only by the ID of a message (the ID can be
determined by the message filename).

=item md5

This method is recommended for low bandwidth operations. This mode can
reduce the file transfers by checking the message moves on the target
side. This mode requires an MD5 sum on the body of the message, so the
first use of this mode can be quite time-consuming on both sides. 

=back

For more information about the algorithms, read the chapter about that.

=item --delete-before, -d

Normally the delete operations on the target side is done after the
transfers. Use this switch if you don't have too much space on your hard
disk. Note that using this switch can reduce the chance of detecting the
message moves.

=item --exclude, -x

Excludes a directory regexp from the transfer and removes it from the
state file. This option can be used more than once to specify more than
one directories. Note, that the directory, which is matched against these
regexps are the relative path of the folders with a leading '/', so if you
want to exclude your Trash folder in the root of your synchronization, then
you have to use the following form:

    --exclude=^/Trash

The excluded paths are used in either source or the target side also. So if
you exclude a very large directory, you will notice speedup in the source
and target side also.

Note that this regexp is matched for every directory that is read from the
filesystem, and every directory what is found in the state file. So if you
provide the exclude pattern as ^/Trash$, then it will skip the Trash
directory when traversing the directory structure, but it WON'T skip files
from the Trash/cur directory when reading from the state file! So be
careful if you use the exclude pattern with existing state file!

=item --exclude-source, --exclude-target

These parameters can be used to exclude files only from one side of the
synchronization. Can be useful with the rename options (below).

=item --rename

Can be used to rename files when transferred from one side to another. A
perl expression is the parameter for this.

If you use this option, use it with care, because you have to provide
exactly the opposite of the name-transformation if you sync to the other
side, for example:

In one side, you can use (A to B):

    --rename="s{^/Saved/}{/ToBeSaved/}" \
    --exclude-target="^/Saved/" \

In the other side, you can use (B to A):

    --exclude-source="^/Saved/" \
    --rename="s{^/ToBeSaved/}{/Saved/}" \

In this case, the Saved folder in the A side will be synchronized with the
ToBeSaved folder in the B side, and the Saved folder in the B side will be
excluded from the synchronization. This scenario can be used when you don't
want to store your emails in the server, but you want to use the "Saved"
folder in the server too. In this case, the emails will be downloaded from
the server (A side) to your laptop (B side), then you can move them to the
Saved folder in B with a script. If this is done, then you can
resynchronize, then the saved files will disappear from the server also.

=back

=head1 ALGORITHMS

Currently the program has two algorithms, which can be used for
synchronization.

=over 4

=item id (default)

This algorithm is based on the message-id-s of the messages. It assumes
that a message-id exists only once in both repositories with the same id.
The id can be determined form the filename.

With this algorithm, you can trace the flag changes or the deletion of a
message and these changes can be propagated to the other side also. It also
handles if a message is copied from the "new" directory to the "cur"
directory without retransmit the files over the network.

This algorithm is recommended if you want a simple and quite fast
operation, and if you have a not-so-slow internet connection.

=item md5

This algorithm does further calculations. It stores the size of the header
and the md5 sum of the message body for each message besides the data that
the "id" algorithm stores.

By using this algorithm, you can track the copies and moves of a message,
so you don't need to retransmit large files if you move them to a new
folder.

If you copy or move a message from one folder to another, the header is
sometimes changed by the mail-reader program. This is why we cannot simply
calculate the MD5 sum of the whole message. A new message in a folder will
have a new identifier also, so it doesn't violate the law that a message-id
is unique. 

When you copy or move a message from your INBOX to your "Save" folder (for
example), the new message is analyzed in the source side, header-size and
md5 sum is calculated on the new message, and from the md5-hash, the source
side can tell the target side what messages has the same hash-value, so the
target side can copy the body from the other message. If the target side has
successfully copied the body from one of those provided messages, then only
the header needs to be transmitted across the network. If the target-side
did not find the messages, then it requests for the body also.

=back

=head1 ONLINE OPERATION

Online operation means that the software can be used in an online mailbox
also. It assumes that the Maildir folder can be changed when the program is
working, so it tries to be as fail-safe as can be.

Every new file is opened in the "tmp" directory, and moved to the
target place only when the file is fully downloaded.

This mode of operation was the first priority, because this feature is
missing from most synchronizer software, including my "drsync" utility
also.

=head1 SPEED

I am using this program to synchronize my Mailboxes. I have 9700 emails in
my mailbox and the state file (bzipped) is 283K.

The first time of a two-way synchronization between a P166 server and a
PIII/1200 notebook over a Cable network, where the starting position is an
already synchronized directory, tooks about 10 minutes. This time is used
for md5-calculation and message-id propagation.

The next two-way run tooks about 40 seconds.

These things are measured in Debian GNU/Linux testing/unstable operating
system (08 Oct 2002).

These are only the overhead of the software, not the real transfer. If you
got a very big email, it needs to be transferred at least one time on the
network. But if you have it in both sides, then it does not require any
more transfer if you save it to different folders.

=head1 TODO

I am currently happy with this feature set, but if I have time, I will
implement these features into the software. Anyway if you have time and
willingness, I accept patches also:

=over 4

=item *

Message-size limit. By limiting the maximum transmitted message, you can
effectively use this software if you have very low bandwidth.

=item *

Config-file and cron-safe operation.

=back

=head1 COPYRIGHT

Copyright (c) 2000-2010 Szab, Balzs (dLux)

All rights reserved. This program is free software; you can redistribute it
and/or modify it under the same terms as Perl itself.

=head1 AUTHOR

dLux (Szab, Balzs) <dlux@dlux.hu>