mddiff [--max-mailno mno] [--db-file dbf] [-v|--verbose] [-d|--dry-run] [--help] paths
If paths is a single file name, and that file is a fifo, mddiff reads from it file names separated by new line and outputs the sha1 of its header and body separated by space.
$ mddiff /tmp/fifo_for_mddiff 806a0ffe4f29766effd764... 463e543da9dac8e298... 582cbb6a5cd3ce13965c8c... 8fa60a7458b1157193... ...If paths is a list of directories, mddiff outputs a list of actions a client has to perform to synchronize a copy of the same maildirs. This set of actions is relative to a previous status of the maildir stored in the db file. The input directories are traversed recursively, and every file encountered inside directories named cur/ and new/ is a potential mail message (if it contains no \n\n it is skipped).
$ mddiff ~/Mail/ ADD ~/Mail/cur/1239038050.14937_1.garfield:2,S 66532ebb05b252e... ...Every client (endpoint using mddiff for synchronization) must use a different db-file, and the db-file is strictly related with the set of directories given as arguments, and should not be used with a different directory set. Adding items to the directory set is safe, while removing them may not do what you want (delete actions are generated).
mddiff does not alter the dbf file, it generates a new one called dbf.new. It is up to the higher level tool smd-server(1) to rename dbf.new to dbf in case the other endpoint successfully applied the diff.