[dev] IMP implements a feature "Request Delivery Confirmation"

[Michael M Slusarz]

Quoting Thomas <thomas at mebes.info>:

> Dear IMP developers,
>
> IMP implements a feature "Request Delivery Confirmation", which is not
> very useful the way it is present. It is also documented in the
> sourcecode of the framework, that the way this is implemented it only
> works with sendmail prior to version 8.7.
> Later versions of sendmail and other mailclients (courier, exim, ...)
> implement the delivery notifications like they are specified in RFC 3461
> (RFC 1891).
> Would it be possible to use this method to request the notifications?
> They do not use an additional RFC2822-Header but pass the request in the
> SMTP envelope. For IMP it would mean, that if it transmits the message
> by a call to /usr/sbin/sendmail it would have to pass an extra argument
> "-N success,delay,fail" to sendmail and if SMTP is used to transmit the
> message, IMP would have to check if the DSN extension is advertized by
> the server and would have to postfix the "RCPT To:<destination at example.com"
> command with " NOTIFY=SUCCESS,DELAY,FAILURE".

I believe this request would be better served if directed toward the 
maintainers
of the PEAR Mail:: class (of which Chuck and Jon are 2) - since we don't
directly call the SMTP servers (we use Mail_smtp::), we can not pass any
additional options to the server unless Mail_smtp:: allows us to (it doesn't
appear to).  We could add it to the sendmail driver (by allowing the user to
specify the additional command line argument in conf.php), but this would
result in some ugly hacks in the IMP/Horde code that would only allow this
preference to be available if using sendmail, etc. so it is not a reasonable
thing to implement at this point.  If the PEAR Mail:: class in the future
provides a clean API to allow us to request the MTA to demand notification,
that would be great.  I'm sure if you provided patches to the
Mail::maintainers, they would be more than happy to take a look.

So until this change happens, we do the only thing possible to request headers
that we can control, which is to try to demand return receipts in the Headers.
I do realize it is very hackish, but it appears to work well in many (or at
least some) situations.

michael

______________________________________________
Michael Slusarz [slusarz at bigworm.colorado.edu]
The University of Colorado at Boulder