Re: fetchmail/exim chokes on a message...
On Thu, Oct 16, 2003 at 02:47:59PM -0400, Derrick 'dman' Hudson (firstname.lastname@example.org) wrote:
> On Thu, Oct 16, 2003 at 10:47:01AM -0700, Sanchez the Cactus wrote:
> | I'm currently using fetchmail set up as a daemon to gather mail from
> | various email accounts and deliver them on my system. default exim3
> | is installed, and users have .procmailrc files to direct mails to a
> | local Maildir.
> | occasionally, i'll stop getting any email, and the syslog will
> | contain a line like:
> | Oct 16 10:24:16 debian fetchmail: message delimiter found while scanning
> | headers
> | Oct 16 10:24:16 debian fetchmail: SMTP error: 501
> | <email@example.com@pop-sbc-v1.mail.vip.sc5.yahoo.com>: malformed
> | address: @pop-sbc-v1.mail.vip.sc5.yahoo.c may not follow
> | <firstname.lastname@example.org
> Junk mail. You told exim to reject mail that is syntactically
> invalid. (syntactically invalid mail only comes from spammers, MS
> OutHouse (when Bccing except under certain circumstances), and some
> bady misconfigured systems (which are broken and need to be fixed
> anyways). You also told fetchmail to send mail to exim via the SMTP
> protocol. The problem is that fetchmail operates after mail has been
> delivered, and fetchmail is not an MTA. So it doesn't handle SMTP
> rejects in an optimal manner because it is a rather unorthodox usage
> of SMTP.
> (FYI exim's logs will contain additional information;
... [more stuff here]
> One solution, one I would recommend anyways, is to add
> with mda '/usr/sbin/sendmail %T'
> to each of those stanzas. That way fetchmail will use the local pipe
> interface instead of SMTP inject the mail into your MTA (exim). This
> interface is simpler, more robust, and not prone to errors like the
I'm having similar problems with my account too, and I was wondering if
using a pipe is the best solution. Wouldnt this mean it would start a
new process for every message it recieves, or at the lease every
account? Is changing the exim option for strict headers better?