On Fri, Feb 13, 2004 at 01:31:00PM -0600, Dan Muey wrote:
[ > I wrote, but Dan snipped the attributions: ]
> > This is almost never the right thing to do.
> I guess I see it as an advantage because you are shortening
> the process and possible places it can break.
And adding in some extras of your own.
> SMTP with whatever language you want :) -> Exim
ECONNREFUSED
ETIMEDOUT
ECONNRESET
ENETDOWN
ENETUNREACH
> Pipe with whatever language (is the language/script configured right?
> IE path to sendmail, sendmail options, permissions, etc...)->
> sendmail (is it linked to Exim properly? Etc Etc -> Exim
If /usr/sbin/sendmail isn't linked to exim properly, you've got more
problems than that, IMO.
> It seems piping from one program to another program that sends
> it to another program has many more points of failure than an SMTP
> session with Exim directly. Regardless of what languages are
> used in what steps.
Why do you have two pipes set up? When you pipe to /usr/sbin/sendmail
you're letting exim take care of the message for you. That's one programme
piping to another. I'm not quite sure where you got your second step.
I'd say it has many less points of failure, actually.
> Maybe I'm missing something?
> > Consider that SMTP is NOT JUST the wire protocol.
You've nicely quoted that, and yet you seem to have completely ignored it.
Perhaps you ought to think about what I might mean here. The current
standard for SMTP is RFC821, but we'll go with 2821 just for good measure.
The sections you want to read are:
4.5.4, 6, 7, 3.1, 3.6, 3.9, and of course, you need to make sure the headers
that you submit are correct, such as a non-reuseable messageid, the date, and
other things.
So, does your client library conform to all of this?
Cheers
MBM
--
Matthew Byng-Maddick <mbm@???> http://colondot.net/
