Postfix 2.1 Released

MasTRE writes "After an extended period of polishing and testing, Postfix 2.1 is released. Some highlights: complete documentation rewrite (long overdue!), policy delegation to external code, real-time content filtering _before_ mail is accepted (a top 10 most requested feature in previous versions), major revision of the LDAP/MySQL/PGSQL code. Version 2.2 is in thw works, which promises even more features like client rate limiting and integration of the TLS and IPv6 patches into the official tree. There's never been a better time to migrate from Sendmail (just _had_ to get that in there ;)."

Comments

Aaargghhh!

[gowen]

It would be nice if, during product announcements, if the submitter actually included a sentence SAYING WHAT THE SOFTWARE DOES.

Yes, I know its an SMTP server, but sheesh, is it so hard to start it "After an extended period of polishing and testing, Postfix, the popular open source mail transfer agent, has reached version 2.1

As my head explodes....

[lacrymology.com]

Wait, wouldn't post fix Postfix 2.1 actually be fix 2.2?

-m

Re:Next time...

[Anonymous Coward]

Many of us are happy with Sendmail

.. as are the kiddies that've r00ted your mail server.

because it's an ugly, lumbering dinosaur

[SuperBanana]

I've been running sendmail 4ever - sure it's complicated as hell - and a bit of a resource hog at times..but it freaking works and is rock solid over more years of production use than any other MTA ever will be in our lifetimes.

On a Axil 320(110mhz, I think? I forget which sparc chip) running Solaris w/320MB of ram and one single SCSI drive, on a Mailman list with about 2,000 subscribers and 100 posts a day, we went from delivery times of an hour+(and load averages well over 4) to under 5 minutes(and load averages between .5 and 2).

Proponents of Sendmail will say "oh, it just needs to be tuned properly".

Nope, sorry. Proper software doesn't need tuning to do its job. Ever notice that the only proponents of the "it just needs someone who knows how to tune it" model are...well...the limited number of people who know how to tune it, and are fast finding themselves out of jobs?

Re:because it's an ugly, lumbering dinosaur

[Just Some Guy]

Proper software doesn't need tuning to do its job.

You may or may not be correct in this particular case, but as a general statement, that's just stupid.

Do you really mean that the exact same settings for a little desktop (high priority to input-related tasks, swap only when needed) would work well for a high-load server (high priority to compute-related tasks, swap agressively to make RAM quickly available)? There are a lot of settings on a modern system that just can't be inferred by the system itself. Stating the opposite like it's an obvious fact is ignorant, misleading, or both.

A real-world example: a Usenet spool and an MP3 repository may be the same size, but benefit hugely from tweaked bytes-per-inode or journal settings. In some cases, once the system is running, it's too late to easily change your mind (like bytes-per-inode). In other cases, you can switch at will, but not without unmounting the filesystem (ext3 journaling options). You, as the administrator, make those decisions. Either way, even if the computer were capable of recognizing that you'd made a bad decision, it's not in a position to correct them.

A real-world example: I tuned Sendmail to use delayed sending so that when a client blasted 20,000 copies of a newsletter (yes, opt-in), then it would wait for several minutes so that it could efficiently aggregate recipients by domain. In there situation, telling Sendmail to leave email in the queue for 10 minutes meant a 50% savings in bandwidth. How on earth would you expect a self-tuned MTA to ever make that discovery on its own?

Computers do some things well. Predicting the future usage patterns of their owners without mounds of previous input is not one of them. That's where manual tuning comes in, and why real system administrators still paid decently.

Postfix Heaven

[Chromodromic]

I just finished installing and configuring Postfix with TLS, Cyrus SASL, Maildir storage (which Postfix simply "does" by appending a "/" at the end of a mailbox path), and virtual users alongside Courier-IMAP, and, man, was it easy. I had the help of O'Reilly's Postfix: The Definitive Guide and between that, the provided documentation and the wealth of resources available on the Web, I was able to get everything up and running in record time.

I know this sounds like a commercial, but it's hard not to sound that way when everything just kind've worked the first time. I now have authenticated, encrypted SMTP and POP and my users are, literally, thanking me. My experience has been that using Postfix was an easy way for me to look good.

Here's a Postfix SASL HOWTO which came in handy, but there are a lot of resources on the Web, especially at the Postfix site.

Real-time filtering

[DustMagnet]

Cool, what's that about? I found this written by Wietse Venema the author/maintianer for postfix:

When used with a real-time SPAM filter, this approach allows Postfix to reject mail before the SMTP mail transfer completes, so that Postfix does not have to return rejected mail to the sender. Mail that is not accepted remains the responsibility of the client.

In all other respects this content filtering approach is inferior to the existing content filter (see FILTER_README) which processes mail AFTER it is queued.

The problem with real-time content filtering is that the remote SMTP client expects an SMTP reply within a deadline. As the system load increases, fewer and fewer CPU cycles remain available to answer within the deadline, and eventually you either have to stop accepting mail or you have to accept unfiltered mail.

Too bad it doesn't have a counter attack mode, yet.

Sendmail upgrade?

[Anonymous Coward]

There's never been a better time to migrate from Sendmail

It seems Exim 4 was released Feb 2002. It includes IPV6, TLS, and SMTPAUTH via PAM, LDAP, MYSQL, PostgreSQL and more.. There is also client rate limiting, and realtime spam/virus filtering no need to accept and bounce junk.
If you're using Postfix and have been waiting for any of these "new features", go ahead and try Exim.
Exim home page

SMTP time scanning, finally.

[stevenbdjr]

real-time content filtering _before_ mail is accepted

About time. I've been doing this with Exim and Exiscan for almost 2 years now. It's nice to see other MTA's begin to incorporate this functionality. Now, if everyone upgrades and takes advantage of this wonderful feature, maybe the number of false NDR's I receive due to forged senders will start to go down...

The Doc

[anarcat]

Yeah, that's good. I always had trouble finding my way into the postfix documentation, now it's a lot clearer. I especially like the listing of all main.cf settings (now if there would be a manpage for master.cf too...) and the bottleneck analysis tool.

I do miss however the "big pictures" yellow + blue graphs that seduced me into trying out postfix long time ago. Now we're stuck with pityful text-only rendering

Still great, after all those years, postfix is my MTA of choice: ease of use, power and security.

Re:to update or not to update?

[arcanumas]

dude, with 8 users you could use trained pigeons and not see a difference.

Re:Comparisons

[Ryquir]

Yeah your comparisons link is seriously outdated (cicra 2001) and only compares mta descriptions. It is neither indepth nor does it touch on the features that existed at the time. With statements like "Add to this sendmail's renowned inefficiency" or "Postfix is quite flexible in its configuration file, but not to the extent of Exim" this document can't be anything more then a abstract draft written up for basic filler in attempt to sell a book idea to publishers.

This wouldn't have been a good comparison at the time it was written let alone now. Next time try googling a little harder perhaps you would have found this link: http://www.geocities.com/mailsoftware42/ or heck google it for yourself here http://www.google.com/search?hl=en&ie=UTF-8&oe=UTF -8&q=MTA+comparison&btnG=Google+Search

Re:Converting from sendmail?

[bearl]

In the source directory there's a text file named INSTALL that has detailed instructions for the three installation options, including "Replace sendmail altogether."

I won't quote them here in case some of the steps have changed, but it's a very nice step by step list of what to do, what to type, and when to type it.

Re:this SMTP server vs Qmail and Sendmail

[DavidTC]

qmail isn't free software because it's non-forkable.

You can freely redistribute the source and binaries compiled from clean source. And you can distribute patches to it.

However, the point is, the qmail maintainer is the only person who can release new versions of qmail. And hence it's not free software.

There are two very large dangers with qmail...that it will go off in a random direction no one agrees with, and you'll either have to follow along or go that way, and that the qmail maintainer will just stop releasing new versions. With free software, if enough people use it, they will simply make a fork...but they can't do that with qmail. Technically they could grab a random version and keep building patches off that, but that becomes unmaintainable real fast.

In other words, qmail is basically 'freeware', not 'free software', although it does come in source form, and you have been granted the ability to modify it and even share the modifications. But not the end result.

Postfix's new policy server API

[RonBurk]

One of the geek-cool things about this release of Postfix is that it finally provides a way to add your own code to the SMTP conversation without having to understand or patch Postfix at all.

The new policy server interface is a simple sockets-based API for getting a chance to participate in the SMTP conversation as it is happening. The basic idea is:

• tell your Postfix config file (main.cf) that you've written a "policy server" that listens on a particular Unix socket or TCP address/port. You can have the policy server get "called" at any of the points in the SMTP conversation where Postfix may make a decision about how to dispose of the message (HELO, RCPT, etc.).
• write your policy server. It listens for connections, and each connection sends you one or more requests. Each request contains a small set of information about the mail message being transmitted (client name/address, HELO text, etc.) Your server responds with one of a broad set of actions that Postfix supports (reject, accept, defer, perform other custom checks, etc.).
• The protocol for talking to your server is a simple text-based protocol with newlines, much like the form of HTTP. I coded an initial policy server in good ol' C in about an hour.

In particular, this new API is a great place to implement greylisting. Your server can maintain its database of whitelisted and greylisted from/to/IP triplets all on its own and just respond to Postfix requests. And, once you've coded up your policy server, you don't have to revise it with every Postfix patch that comes down the pike. As long as the API remains backwardly compatible, your policy server code should survive any Postfix upgrades.

Kudos to the new policy server API!

When next you announce..

[Anonymous Coward]

The latest version of an application... how about including a link to the release notes / changelog. No point in upgrading if you don't know the changes - RELEASE_NOTES

Re:this SMTP server vs Qmail and Sendmail

[kweber666]

There are two very large dangers with qmail...that it will go off in a random direction no one agrees with

There is another theory which states that this has already happened.

and that the qmail maintainer will just stop releasing new versions

To quote the qmail web site: The latest published qmail package is qmail-1.03.tar.gz, which was released in June 1998. So again, this may have happened already.