[GLLUG] Postfix or Exim?

Matt Graham danceswithcrows at usa.net
Thu Jan 11 16:26:30 EST 2007


On Thursday 11 January 2007 14:22, after a long battle with technology, 
Ian Fitzpatrick wrote:
> wondering what people's thoughts are on Postfix vs. Exim.  I've used
> both, but only on testing or home systems.

> 1.  Stable and secure
> 2.  Plays nice with dovecot
> 3.  Ease of administration (initial configuration not withstanding)

I have exim+spamassassin running for a small business.  Postfix was 
easier to configure, but exim appears to be a lot more flexible in 
terms of the things you can do.  It should be possible to have exim 
deliver mail to ~/maildir/ instead of /var/mail/$USER with a simple 
directive that's published in the README that comes with exim.  I don't 
remember how you do that with Postfix.

> 5.  Marketability:  This is a bad reason to pick technology I know,
> but I am kind of curious which of these are most widely deployed

qmail may be more widely deployed than postfix or exim, actually.  Not 
sure though, just saying based on what the non-work mail provider I use 
uses and the mail/webhosting place that we were forced to use for a 
while uses.

> I read some comparison articles on the two MTAs, but
> it usually seems like it's a wash.  What might tip you in favor of
> one versus the other?

Sane error messages in the logfiles, reasonably easy/automated stats 
collection, minimal screwing around needed once it's set up, and a 
config file that doesn't look like an explosion in an EBCDIC factory.  
That's just me though.


Scott Harrison wrote in another message that showed up while I was 
already writing this one:
> http://yro.slashdot.org/article.pl?sid=06/12/01/1425221&tid=187 

I'd say that ain't enforceable, but I pity the poor people who might 
have to test that in court.  Damn stupid lawyers/judges.

> I would also wonder about the business economics in terms of 
> maintaining an e-mail server at 100% over the long-run given the 
> vagaries of spam, etc. 

IME, if you have a reasonably competent tech in house, you'd be *better 
off* having mail on a Linux/*BSD box in the server rack.

Dec 2005: old mailserver falls over due to hardware failures and old 
age.  I resurrect it with much cursing, kicking of things, and 
replacing a SCSI card, and spec out a reasonable, inexpensive box to 
replace it.

Jan 2006: new box gets here.  I spend 3 days installing/configging 
everything and running a bunch of tests to make sure everything will 
work right when the switchover occurs.  Switch happens, only 1 problem 
that's quickly fixed, users happy.

Then CEO decides that we are too small of a company to need an in-house 
mailserver, arranges hosting through some guys he knows, and transfers 
the MX record.  All of this is done without my knowledge.  Then 
people's usernames are wrong.  People's passwords are incorrect.  
Aliases aren't enabled.  People can't receive individual messages > 10M 
because of quotas.  (Yes, we have legit traffic larger than that; 
customers who can't use FTP/SCP mailing enormous TIFFs and PDFs and so 
forth.)  Sending large files from 1 person to another in the same 
office via mail takes forever, as mail must go from office->rack 
somewhere in Detroit->office, twice, with the slowest link in the chain 
being 2 T1s.

Dealing with this monumentally stupid decision cost me and the users a 
great deal of time.  As the CEO was "trying to save the tech staff's 
time", this is exactly what you'd expect from someone who doesn't 
understand technology and didn't ask the people who *did* understand it 
what was needed.  The fix was simple--register new domain name, point 
MX record at old mailserver, tweak mailserver's config a bit, tell 
everybody to use newdomain.com, and everything worked *fine*.

> But if your organization insists on having its own in-house solution, 
> maybe it makes sense to get a packaged solution as opposed to a 
> build-your-own? (http://www.open-mag.com/1823548701.shtml)

"apt-get install exim4-daemon-heavy" is packaged, shirley?  And it's 
less expensive, lots more expandable, and probably less likely to break 
than the $600 thing those guys are trying to sell.  People who don't 
know anything about tech are the audience for that thing.  The fact 
that Ian's asking the questions he's asking shows he's not in that 
category and will be ill-served by the products marketed to that 
category.

-- 
   A project for developing fully automatically driven cars
   (not just those misnamed auto-pilots) has been improved by your
   explosive flatulence.  --MegaHAL, trained on ASR
There is no Darkness in Eternity/But only Light too dim for us to see


More information about the linux-user mailing list