This documentation is for Dovecot v1.x, see wiki2 for v2.x documentation.

MH Mailbox Format

The MH mailbox format originated with a system developed by the RAND corporation and the University of California. Each email message is stored in a single file, with directories indicating folders and subfolders. The index or order of the messages in the folder determine what each message is named (which may not correspond to the inode index). The "safe" way to guarantee a message gets written to a mail folder is to first write the message out to a randomly chosen temporary file name, then link or rename the file to the number LAST+1, where LAST is the last sequential message in the folder. If the rename fails, increment the counter and try again.

MH folders also maintains a meta-file called ~/Mail/.mh_context that contains information about the most current folder and message chosen. Each sub-folder also contains a meta-file called .mh_sequences or .xmhcache, which maintains keyword association lists for stored queries. New messages are stored in the "unseen" sequence for a folder. Procmail itself does not bother making changes to this file, rather simply delivers the message to the folder and leaves determining new messages as an exercise for the MUA. For example:

unseen: 1-3 8 15
projectB: 2-8 10

shows two stored sequences of messages. Command-line utilities can then use these sequences as shortcuts. show unseen, for example, is short-hand for show 1-3 8 15.

Deleted emails are indicated by prepending a "," to the name. One of the largest problems that IMAP servers have with MH format is the volatility of the email message name itself. The command-line utility sortm is used to sort mail folders by date or string matching. To do this, messages are actually renamed to reflect the new sort order. IMAP servers are required to maintain an index of the folder contents, so when the names of the file entries cannot be guaranteed to be stable, IMAP servers have to throw out previous index caches and re-index.

When operating with a shell account on a machine that also provides IMAP access to folders, users are encouraged not to re-sort email locally while accessing the IMAP server remotely.


None: MailboxFormat/MH (last edited 2009-03-15 22:35:09 by localhost)