Index: [thread] [date] [author] [stats]
  From: markus schnalke <meillo@marmaro.de>
  To  : <masqmail@marmaro.de>
  Date: Mon, 01 Nov 2010 18:06:25 +0100

-t option and cmdline arguments

Hoi,

the -t option tells masqmail to read the recipients from the mail
headers. (Usually the headers are not looked at but the recipients are
taken on the command line.)

The case of question is what to do if -t is given *and* addresses on
the command line. Here the other MTAs behave differently:

- smail, exim, masqmail (all the same family): substract the cmdline
recipients from the recipients in the headers.

- postfix: add the cmdline recipients to the recipients in the
headers, which is the contrary. (Earlier versions of postfix failed
if -t was given together with recipients on the cmdline.)

- sendmail: behaved differently, depending on the version.


I dislike the current behavior as it feels to be complicated
(especially in the want for clear rules) and seems to be of no
practical value. In general, the whole corner case of cmdline
addresses together with -t is useless as different MTAs behave
differently. But failing in such a situation, as postfix once did,
seems to be bad because we need to maintain sendmail-compatibility.
(I had this problem just now with the unrecognized -m option.) This
probably is the reason why postfix does not fail in this situation
anymore.

As neither substracting nor adding the cmdline addresses seems to be
the right behavior, I suggest to simply ignore them. The idea behind
this is seeing -t as ``read recipient addresses from headers *instead*
of from the cmdline arguments''. As masqmail doesn't look at the
headers in the normal (without -t) case, it shouldn't look at the
addresses given on the cmdlin in the -t case.


What do you think?

Can you imagine scenarios where the current behavior in this corner
case is important to keep, though other MTAs behave differently
anyway?


meillo


Index: [thread] [date] [author] [stats]