| One solution is to replace portions of the logged text by [...]
| and to display the portion of text that contains the offending data.
| reject: <error-class>: <description of error>:\
| Subject: humpty dumpty [...] together again
| But this would be solving the wrong problem.
|
| If you want to know what mail gets caught in your nets, then you
| must not reject it, but you must file it for later inspection.

For 90% of my rejected cases the cause is one (or a few)
unencoded national 8-bit characters placed either in the Subject,
the comment part of the From or To address,
or some silly local reseller tried to localize some product, e.g.:

  X-Mailer: Microsoftova internetna e-po?ta/MAPI

And, ouch, I have seen localized date/time in the 'Date:' header!

In all these common case there is no need to accept and inspect
the mail, a simple reject with logged problem line suffices.

The '?' replacement is good enough for my purpose, although
the \nnn or %hh would be more informative. A long formatted string
could be chopped off or portions replaced with [...],
as long as the logged portion includes at least the first
problem character.

Who knows, one may even capture interesting control characters
( e.g. some \0 or <CR> resulting from some broken SMTP session :)
which would not be apparent from the '?' form.

  Mark
-
To unsubscribe, send mail to majordomopostfix.org with content
(not subject): unsubscribe postfix-users

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]