[linux-elitists] Postfix vs Exim

Jeff Waugh jdub@perkypants.org
Sat Sep 21 22:57:28 PDT 2002


<quote who="Marc MERLIN">

> > > - I know  that because postfix isn't  monolithic like exim, and  that one of
> > >   the prices to pay is that it doesn't do deduping (i.e. if a message is
> > >   sent to you and an email alias, you get two copies). If postfix's secure
> > >   module approach has other drawbacks, I'm interested to hear them
> > 
> > I'd call this a very peripheral feature for an *MTA*. You can do it in
> > procmail or some MUAs.
>  
> Right, but  it doesn't  work if you  feed the message  directly to  Cyrus or
> similar, and not that many MUAs allow deduping by message ID
> Unfortunately, that was almost the single reason why postfix didn't replace
> sendmail at google for instance.

I just wouldn't be doing this in my MTA. I could let an MDA or another
transport do it, but it doesn't sound like a critical core feature.

> > > - I don't think postfix can do SMTP callbacks or callouts like exim.
> > >   Confirmation appreciated.
> > 
> > Non-exim-jargon feature description, please.
> 
> Sorry. Description here:
> http://sourceforge.net/docman/display_doc.php?docid=6747&group_id=1 

Hmm, it does every check up to this, not entirely sure if it will actually
attempt to connect and validate the email delivery though... Could be wrong.

> > > - I'm not sure that postfix can do the equivalent of exim's local scan or
> > >   sendmail's milters. Confirmation appreciated
> > 
> > Apart from the header/content checks, it supports external transports and
> > transport tables. This is far, far more scalable than milters and such.
> > Rather than launching extra processes, etc., you send the mail to another
> > transport via pipe, lmtp, smtp, yada.
> 
> Well, exim does this too :-)  but the idea of a milter or exim's local scan
> is that you do it at SMTP time.
> What you describe is  done after the mail has been  accepted, which is often
> too late to bounce it.

I've been happy with the number of checks that postfix already offers
through the receive process. As well as the header/body checks, it now has a
data restrictions hook, but thus far I've only seen it used for rejecting
pipelining at that stage.

I'm sure Wietse has answered this a hundred times or more though, I'll peek
in the archives to see.

> > > - postfix has an extensive lookup cache, which is a huge advantage over exim
> > >   if you are doing  ldap lookups (ldap is slow, and exim  will not cache the
> > >   lookup longer than the process runs)
> > 
> > You might want to research that a bit more, it's a religious issue. :-)
>  
> What do  you mean? I  checked that  exim won't  cache beyond  the life  of a
> subprocess and a friend  who knows postfix well told me  the caching he used
> to do with LDAP on postfix.
> Are you saying that the caching isn't very good or that caching is bad?

Caching can be bad, and there have been issues with client-side caching with
OpenLDAP, so there's a lot of FUD about it. There are definite plans to make
caching better with postfix though (sounded really cool last time I read
about it).

> > > - Does postfix  support complete header  rewriting (on all the  headers, per
> > >   header and with regular expressions), like exim does?
> > 
> > No. (Why would an MTA need to do this, if it couldn't replace the
> > configurability with a sensible feature?)
>  
> So you're  saying that  you'd need to  tell postfix to  pipe the  program to
> something else that does all the rewriting?

No, I'm wondering why you'd do header rewriting manually in the first place,
rather than adding a system to the MTA to handle common requirements for you
with sensible configuration options.

- Jeff

-- 
   "She sped into the parking lot, the door flew open and there she was,    
   dressed only in a black bathing suit and sunglasses. I got in and she    
     reversed straight into another car. Classic Carine." - Stephen Gan     



More information about the linux-elitists mailing list