The stable Postfix release is called postfix-2.2.x where 2=major release number, 2=minor release number, x=patchlevel. The stable release never changes except for patches that address bugs or emergencies. Patches change the patchlevel and the release date. New features are developed in snapshot releases. These are called postfix-2.3-yyyymmdd where yyyymmdd is the release date (yyyy=year, mm=month, dd=day). Patches are never issued for snapshot releases; instead, a new snapshot is released. The mail_release_date configuration parameter (format: yyyymmdd) specifies the release date of a stable release or snapshot release. Incompatibility with Postfix 2.1 and earlier ============================================ If you upgrade from Postfix 2.1 or earlier, read RELEASE_NOTES-2.2 before proceeding. Major changes with snapshot 20051215 ==================================== Plug-in support for SASL authentication in the SMTP server. The default plug-in is Cyrus SASL, so everything should behave like it did before. The SMTP/LMTP client still needs to be made pluggable. Incompatibility with snapshot 20051208 ====================================== The fallback_relay feature is renamed to smtp_fallback_relay, to make clear that the combined SMTP+LMTP client uses this setting only for SMTP deliveries. The old name still works. The LMTP client now reports the server as "myhostname[/path/name]". With the real server hostname in delivery status reports, the information will be more useful. Major changes with snapshot 20051208 ==================================== The SMTP client now implements the LMTP protocol. Most but not all smtp_xxx parameters have an lmtp_xxx "ghost" parameter. This means there are lot of new LMTP features, including support for TLS and for the shared connection cache. There are no lmtp_xxx "ghost" parameters for the HELO or EHLO commands, because those commands exist only in SMTP. Incompatibility with snapshot 20051202 ====================================== The Postfix SMTP daemon will not receive mail from the network if it isn't running with postfix mail_owner privileges. This prevents surprises when, for example, "sendmail -bs" is configured to run as root from xinetd. Incompatibility with snapshot 20051125 ====================================== You MUST stop and restart Postfix, because the address resolver protocol has changed. If you don't stop and restart Postfix, you will have an endless stream of warning messages with "problem talking to service rewrite: Unknown error: 0" and "warning: unexpected attribute address in input from rewrite socket". Major changes with snapshot 20051125 ==================================== This snapshot adds support for sender-dependent ISP accounts. - Sender-dependent smarthost lookup tables. The maps are searched with the sender address and with the sender @domain. The result overrides the global relayhost setting, but otherwise has identical behavior. See the postconf(5) manual page for more details. Example: /etc/postfix/main.cf: sender_dependent_relayhost_maps = hash:/etc/postfix/sender_relay - Sender-dependent SASL authentication support. This disables SMTP connection caching to ensure that mail from different senders will use the correct authentication credentials. The SMTP SASL password file is first searched by sender address, and then by the remote domain and hostname as usual. Example: /etc/postfix/main.cf: smtp_sasl_auth_enable = yes smtp_sender_dependent_authentication = yes smtp_sasl_password_maps = hash:/etc/postfix/sasl_pass Incompatibility with snapshot 20051121 ====================================== For compatibility reasons, the permit_mx_backup feature will accept mail for authorized destinations (see permit_mx_backup for definition). Only with other destinations will it require that the local MTA is listed as non-primary MX. Incompatibility with snapshot 20051120 ====================================== The permit_mx_backup feature now requires that the local MTA is not listed as primary MX host for the recipient domain. This prevents mail loop problems when someone points the primary MX record at Postfix. Major changes with snapshot 20051113 ==================================== Configurable bounce messages, based on a format that was developed by Nicolas Riendeau. The file with templates is specified with the bounce_template_file parameter. Details are in the bounce(5) manual page, and examples of the built-in templates can be found in a file bounce.cf.default in the Postfix configuration directory. The template for the default bounce message looks like this: failure_template = < If you do so, please include this problem report. You can delete your own text from the attached returned message. The $mail_name program EOF Incompatibility with snapshot 20051106 ====================================== The relay=... logging has changed and now includes the remote SMTP server port number as hostname[hostaddr]:port. Incompatibility with snapshot 20051105 ====================================== qshape needs to be updated. See the file qshape-microsecond-patch. All delay logging now has sub-second resolution, including the over-all "delay=nnn" logging. At this point the Postfix logging for a recipient looks like this: Nov 3 16:04:31 myname postfix/smtp[30840]: 19B6B2900FE: to=, orig_to=, relay=mail.example.com[1.2.3.4], conn_use=2, delay=0.22, delays=0.04/0.01/0.05/0.1, dsn=2.0.0, status=sent (250 2.0.0 Ok) Incompatibility with snapshot 20051103 ====================================== pflogsumm needs to be updated. See the pflogsumm-conn-delays-dsn-patch file. The queue manager protocol has changed. You need to "postfix reload" after "make upgrade". The logging of recipient status information has changed. This may require changes to logfile processing tools. - Postfix now logs an additional attribute with detailed delay information (delays=a/b/c/d) as described below. - Postfix now logs an additional attribute with the connection reuse count (conn_use=nnn) as described below. Major changes with snapshot 20051103 ==================================== This release makes a beginning with a series of new attributes in Postfix logfile records. - Better insight into the nature of performance bottle necks, with detailed logging of delays in various stages of message delivery. Postfix logs additional delay information as "delays=a/b/c/d" where a=time before queue manager, including message transmission; b=time in queue manager; c=connection setup time including DNS, HELO and TLS; d=message transmission time. - Logging of the connection reuse count when SMTP connections are used for more than one message delivery. This information is needed because Postfix can now reuse connections hundreds of times or more, and can help to diagnose interoperability problems with servers that suffer from memory leaks or other resource leaks. At this point the Postfix logging for a recipient looks like this: Nov 3 16:04:31 myname postfix/smtp[30840]: 19B6B2900FE: to=, orig_to=, relay=mail.example.com[1.2.3.4], conn_use=2, delay=0, delays=0/0.01/0.05/0.1, dsn=2.0.0, status=sent (250 2.0.0 Ok) The following two logfile fields may or may not be present: orig_to This is omitted when the address did not change. conn_use This is omitted when a connection is used once. Incompatibility with snapshot 20051026 ====================================== The connection cache protocol for SMTP connections has changed. You need to "postfix reload" after "make upgrade". The smtp_connection_cache_reuse_limit parameter (which limits the number of deliveries per SMTP connection) is replaced by the new smtp_connection_reuse_time_limit parameter (the time after which a connection is no longer stored into the connection cache). Major changes with snapshot 20051026 ==================================== This snapshot addresses a performance stability problem with remote SMTP servers. The problem is not specific to Postfix: it can happen when any MTA sends large amounts of SMTP email to a site that has multiple MX hosts. The insight that led to the solution, as well as an initial implementation, are due to Victor Duchovni. The problem starts when one of a set of MX hosts becomes slower than the rest. Even though SMTP clients connect to fast and slow MX hosts with equal probability, the slow MX host ends up with more simultaneous inbound connections than the faster MX hosts, because the slow MX host needs more time to serve each client request. The slow MX host becomes a connection attractor. If one MX host becomes N times slower than the rest, it dominates mail delivery latency unless there are more than N fast MX hosts to counter the effect. And if the number of MX hosts is smaller than N, the mail delivery latency becomes effectively that of the slowest MX host divided by the total number of MX hosts. The solution uses connection caching in a way that differs from Postfix 2.2. By limiting the amount of time during which a connection can be used repeatedly (instead of limiting the number of deliveries over that connection), Postfix not only restores fairness in the distribution of simultaneous connections across a set of MX hosts, it also favors deliveries over connections that perform well, which is exactly what we want. The smtp_connection_reuse_time_limit feature implements the connection reuse time limit as discussed above. It limits the amount of time after which an SMTP connection is no longer stored into the connection cache. The default limit, 300s, can result in a huge number of deliveries over a single connection. This solution will be complete when Postfix logging is updated to include information about the number of times that a connection was used. This information is needed to diagnose inter-operability problems with servers that exhibit bugs when they receive multiple messages over the same connection. Incompatibility with snapshot 20051011 ====================================== The Postfix local(8) delivery agent no longer updates its idea of the Delivered-To: address while it expands aliases or .forward files. With deeply nested aliases or .forward files, this can greatly reduce the number of queue files and cleanup process instances. To get the earlier behavior, specify "frozen_delivered_to = no". The frozen_delivered_to feature can help to alleviate a long-standing problem with multiple deliveries to recipients that are listed multiple times in a hierarchy of nested aliases. For this to work, only the top-level alias should have an owner- alias, and none of the subordinate aliases. Major changes with snapshot 20051011 ==================================== Optional protection against SMTP clients that hammer the server with too many new (i.e. uncached) SMTP-over-TLS sessions. Cached sessions are much less expensive in terms of CPU cycles. Use the smtpd_client_new_tls_session_rate_limit parameter to specify a limit that is at least the inbound client concurrency limit, or else you may deny legitimate service requests. Optional suppression of remote SMTP client hostname lookup and hostname verification. Specify "smtpd_peername_lookup = no" to eliminate DNS lookup latencies, but do so only under extreme conditions, as it makes Postfix logging less informative. Incompatibility with snapshot 20050828 ====================================== When a header/body_checks or message_reject_characters rule rejects mail that was submitted with the Postfix sendmail command (or re-queued with "postsuper -r"), the returned message is now limited to just the message headers, to avoid the risk of exposure to harmful content in the message body or attachments. When the cleanup server rejects the content or size of mail that was submitted with the Postfix sendmail command, forwarded with the local(8) delivery agent, or that was re-queued with "postsuper -r", Postfix no longer sends DSN SUCCESS notification of virtual alias expansions. Since all the recipients are reported as failed, the SUCCESS notification seems redundant. Major changes with snapshot 20050828 ==================================== Configurable filters to reject or remove unwanted characters in email content. The message_reject_characters and message_strip_characters parameters understand the usual C-like escape sequences: \a \b \f \n \r \t \v \ddd (up to three octal digits) and \\. Incompatibility with snapshot 20050726 ====================================== Name server replies that contain a malformed hostname are now flagged as permanent errors instead of transient errors. This change works around a questionable proposal to use syntactically invalid hostnames in MX records. Major changes with snapshot 20050724 ==================================== SMTPD Access control based on the existence of an address->name mapping, with reject_unknown_reverse_client_hostname. There is no corresponding access table lookup feature, because the name is not validated in any way (except that it has proper syntax). Several confusing SMTPD access restrictions were renamed: reject_unknown_client -> reject_unknown_client_hostname, reject_unknown_hostname -> reject_unknown_helo_hostname, reject_invalid_hostname -> reject_invalid_helo_hostname, reject_non_fqdn_hostname -> reject_non_fqdn_helo_hostname. The old names are still recognized and documented. Incompatibility with snapshot 20050716 ====================================== Internal interfaces have changed; this may break third-party patches because the text of function argument and result type definitions has changed. The type of buffer lengths and offsets were changed from "(unsigned) int" (32 bit on 32-bit and LP64 systems) to "(s)size_t" (64 bit on LP64 systems, 32 bit on 32-bit systems). Otherwise, this change makes no difference on 32-bit systems. On LP64 systems, however, software may mis-behave 1) when Postfix is linked with pre-compiled code that was compiled with old Postfix interface definitions and 2) when compiling Postfix source that was modified by a third-party patch: incorrect code may be generated when the patch passes the wrong integer argument type in contexts that disable automatic argument type conversions. Examples of such contexts are formatting with printf-like arguments, and invoking functions that write Postfix request or reply attributes across inter-process communication channels. Unfortunately, gcc does not report "(unsigned) int" versus "(s)size_t" format string argument mis-matches on 32-bit systems; they can be found only on 64-bit systems. Major changes with snapshot 20050716 ==================================== Improved portability to LP64 systems, by converting the type of buffer lengths and offsets from "(unsigned) int" to "(s)size_t". This change has zero effect on 32-bit systems. On LP64 platforms, however, this change not only eliminates some obscure portability bugs, it also eliminates unnecessary conversions between 32/64 bit integer types, because many system library routines take "(s)size_t" arguments or return "(s)size_t" values. Incompatibility with snapshot 20050707 ====================================== The connection cache protocol is changed. You may need to "postfix reload" after upgrading. Incompatibility with snapshot 20050627 ====================================== The Postfix SMTP client no longer applies the smtp_mx_session_limit to non-permanent errors during the TCP, SMTP, HELO or TLS handshake. Previous versions did that only with TCP and SMTP handshake errors. Incompatibility with snapshot 20050622 ====================================== The Postfix SMTP client by default limits the number of MX server addresses to smtp_mx_address_limit=5. Previously this limit was disabled by default. The new limit prevents Postfix from spending lots of time trying to connect to lots of bogus MX servers. The Postfix SMTP error handling of [45]XX server greetings was cleaned up. The server reply is now properly reported. Incompatibility with snapshot 20050615 ====================================== Many internal protocols have changed. You must reload Postfix or else the queue manager and delivery agents will complain about unexpected request and reply attributes. The new DSN support conflicts with VERP support. For Sendmail compatibility, Postfix now uses the sendmail -V command line option for DSN. In order to request VERP style delivery, you must now specify -XV instead of -V. The Postfix sendmail command will recognize if you try to use -V for VERP-style delivery. It will do the right thing and will remind you of the new syntax. The queue file format is backwards compatible (again) with Postfix 2.2. Postfix 2.3 stores attributes that older versions will ignore. Major changes with snapshot 20050615 ==================================== DSN support as described in RFC 3461 .. RFC 3464. This gives senders control over successful and failed delivery notifications. DSN involves extra parameters to the SMTP MAIL FROM and RCPT TO commands, as well as extra Postfix sendmail command line options that provide a sub-set of the functions of those extra SMTP command parameters. See DSN_README for details. Some implementation notes are in DSN_NOTES, in the top-level source code directory. Major changes with snapshot 20050510 ==================================== This release improves usability of enhanced status codes in Postfix access tables, RBL reply templates and in transport maps that use the error(8) delivery agent. - When the SMTP server rejects a sender address, it transforms a recipient DSN status (e.g., 4.1.1-4.1.6) into the corresponding sender DSN status, and vice versa. - When the SMTP server rejects non-address information (such as the HELO command parameter or the client hostname/address), it transforms a sender or recipient DSN status into a generic non-address DSN status (e.g., 4.0.0). These transformations are needed when the same access table or RBL reply template are used for client, helo, sender, or recipient restrictions; or when the same error(8) mailer information is used for both senders and recipients. Incompatibility with snapshot 20050503 ====================================== The format of some "warning:" messages in the maillog has changed so that they are easier to sort: - The logging now talks about "access table", instead of using three different expressions "access table", "access map" and "SMTPD access map" for the same thing. - "non-SMTP command" is now logged BEFORE the client name/address and the offending client input, instead of at the end. Major change with snapshot 20050427+DSN ======================================= This is experimental DSN support added to snapshot 20050427. The code is not for production purposes; it is not fully tested, some names and interfaces are still rough around the edges, and it does not update the oqmgr so you have to use qmgr instead. Some implementation notes and open issues are described in the DSN_SUPPORT_README file (top-level directory). Incompatibility with snapshot 20050329 ====================================== If you use TLS, you need to execute "postfix reload" because the TLS manager protocol has changed. Incompatibility with snapshot 20050328 ====================================== The logging format has changed. Postfix delivery agents now log the RFC 3463 enhanced status code as "dsn=x.y.z" where y and z can be up to three digits each. See the file pfloggsum-dsn-patch for an update to the pfloggsum script. After you upgrade from Postfix 2.2 or 2.3 you need to execute "postfix reload", otherwise you will keep running the old Postfix queue manager, which gives no special treatment to the enhanced status codes that it receives from Postfix delivery agents. Major changes with snapshot 20050328 ==================================== This release introduces support for RFC 3463 enhanced status codes. For example, status code 5.1.1 means "recipient unknown". Postfix recognizes enhanced status codes in remote server replies, generates enhanced status codes while handling email, and reports enhanced status codes in non-delivery notifications. This improves the user interaction with mail clients that hide the text of error messages from users. You can, but don't have to, specify RFC 3463 enhanced status codes in the output from commands that receive mail from a pipe. If a command terminates with non-zero exit status, and an enhanced status code is present at the beginning of the command output, then that status code takes precedence over the non-zero exit status. You can, but don't have to, specify RFC 3463 enhanced status codes in Postfix access maps, header/body_checks REJECT actions, or in RBL replies. For example: REJECT 5.7.1 You can't go here from there The status 5.7.1 means "no authorization, message refused", and is the default for access maps, header/body_checks REJECT actions, and for RBL replies. If you specify your own enhanced status code, the Postfix SMTP server will automatically change a leading '5' digit (hard error) into '4' where appropriate. This is needed, for example, with soft_bounce=yes.