a dot alone in a line (bug?)

[Matthias Andree]

bogo at escom.com writes:

> Yes, it is a bug if a sending MTA sends a message with such a dot 
> embedded in the message itself.  But then the problem would be 
> with the sending MTA.
>
> The "dot on a line by itself" is an RFC standard rather than
> just a mechanism used by sendmail.  See Example 1 of RFC 821,
> which shows a <CR><LF>.<CR><LF> sequence as the end of message.
> The first CRLF is usually generated at the end of the previous (last)
> line, and so the dot indeed begins on a line by itself.

Careful there. This line-stuffing (single dot alone is end-of-message
marker, other dots are doubled) is only the over-the-wire format
(RFC-2821) and not necessarily seen in Postfix integration as was common
for the reported problems. If you chat via commands, it all depends on
whether you use sendmail's "-i" or "-oi" option or not.

> Different implementations may record the dot at the end of message 
> and strip it later on.  I think that's what Matthias is suggesting 
> when he asks for the the code used to interface bogofilter to 
> other mail software.

Exactly that. If the content filter is a pipe(8) based script, I would
like to see the master.cf entry and the script, for instance.

> I don't know what assumptions bogofilter makes or what it does
> when it sees such a dot, but if bogofilter terminates scanning
> and returns the result up to that point, then I would argue that
> bogofilter is doing the best it can with bad data.

Bogofilter doesn't care for SMTP conventions. What matters is RFC-2822,
which has nothing to do with dot stuffing.

-- 
Matthias Andree