Overview of Archiveopteryx.
Archiveopteryx is a mail server system optimised for long-term storage and heavy
access. It comprises a set of server programs that provide access to mail
stored in a relational database (PostgreSQL).
The following five steps should suffice to install Archiveopteryx. If they
don't, or if anything is unclear, more detailed instructions are available at
Usually, this is as simple as:
- apt-get install archiveopteryx (on Debian)
pkg install archiveopteryx (on FreeBSD)
http://archiveopteryx.org/installation explains what to do in other cases.
- /usr/local/libexec/aox/archiveopteryx start
to start Archiveopteryx using the default configuration from
, which tries hard to be sensible.
- aox add user name password address@domain
to create a username and password.
You can use deliver
(8) to inject some mail you already have.
This command injects all the messages from a berkeley mbox file:
- formail -s deliver yourname@domain < mbox
The mail you just delivered is in your inbox, and you can read it with any IMAP
or POP3 client.
The normal way to deliver mail from your MTA into Archiveopteryx is via LMTP. By
default, Archiveopteryx listens on 127.0.0.1 port 2026.
On aox.org we describe how to configure some common MTAs to work with
Archiveopteryx. See e.g. http://archiveopteryx.org/postfix for
Archiveopteryx consists of a number of frontend servers, each of which uses
- The RDBMS is where all the mail is stored. At present only PostgreSQL is
- The log server is an internal server which filters log entries so useful
messages are logged while noise is suppressed.
Each server is described more fully in its own man page: logd
(1) and archiveopteryx
(8) for the frontend server, which
serves IMAP, POP3, LMTP and/or SMTP and perhaps webmail.
Normally, there is one logd
(8) process running. In addition there will be
at least one postgres
(1) and one archiveopteryx
The servers use TCP to communicate internally, so they can be distributed across
a server cluster. Generally, one host will run the database backend, one host
(perhaps the same) will run logd
(8) and as many as required will run
The man pages for each server explain the use and configuration of each. See
(5) for more about configuring Archiveopteryx in
All Archiveopteryx servers run in chroot directories.
The user-facing servers run in a special jail directory. They have neither read
nor write access to this directory.
(8) runs chrooted to another directories. All of the servers close
all open files at startup and drop root privileges. By default they run as
, group aox
, although these names can be changed using
The servers check that they lose prvileges as expected, and refuse to start if
they're too privileged.
Note that logd
(8) must have permission to create the logfile
Archiveopteryx does not store mail in the RFC-822 format. It parses each message
upon delivery, and stores a normalized representation, optimized for fast and
reliable search and categorization.
This offers the following advantages, among others:
- One example: Each address is stored exactly once, as Unicode and with
RFC-2047 encoding undone. Finding all messages sent from/to a given name
is extremely fast, because only a single SQL SELECT is necessary, and it
accesses just two small tables.
- Long-term Stability
- Archiveopteryx handles today's common syntax problems and stores the
correct form in the database. Because of that, the mail reader which looks
at old mail in the year 2020 will not need to be bug compatible with
today's version of Microsoft Outlook.
- Scalability and Flexibility
- Only the database size limits Archiveopteryx's capacity. Many other
servers limit individual folders to 2GB (or less), or cannot support more
than a few thousand subfolders/messages in a folder. With Archiveopteryx,
you don't need to invent workarounds for such artificial limitations.
- Because mail is stored normalized and parsed, large attachments are
generally stored only once, and mail parsing exploits are rendered
harmless before reaching the MUA. A movie clip sent to a hundred
recipients is not a problem, because it's stored just once.
Archiveopteryx is available under the PostgreSQL License.
The configurable file and directory names in this build are as follows:
- (where servers live) is /usr/local/sbin.
- (where other executables live) is /usr/local/bin.
- (where the startup script lives) is /usr/local/etc/rc.d.
- (where manpages live) is /usr/local/man.
- (where pidfiles live) is /var/run/aox.
- (where supporting files live) is /usr/local/libexec/aox.
- (the working directory of the user-facing servers) is
/var/db/aox/jail, and can be overridden using the
jail-directory variable in archiveopteryx.conf(5).
- (where the configuration files live) is
- (the full name of the logfile) is syslog/mail, and can be
overridden using the logfile variable in
These variables can be changed only by editing the file Jamsettings and
recompiling Archiveopteryx. Jamsettings also contains some variables used only
during compilation and/or installation, and some which provide defaults for
- contains the Archiveopteryx configuration.
- contains a private key and self-signed certificate used by
The Archiveopteryx Developers, email@example.com.
This man page covers Archiveopteryx version 3.2.0, released 2014-03-10,