Description of problem:

The sendmail man page describes the -i option as follows:

"-i     Ignore dots alone on lines by themselves in incoming
messages.  This should be set if you are reading data from a file."

In reality, sendmail ordinarily performs the "dot-stuffing"
behavior described in Section 4.5.2 of RFC2821 when it reads
a message on standard input.  That is, it usually removes
a leading period on a line (irrespective of whether the line
has additional characters).  When given the -i option, it does
not strip leading periods.  Part of the result is that -i
causes it to "ignore dots alone on lines by themselves in incoming
messages," but this flag also affects sendmail's behavior on all
lines that begin with dots, even when they are not alone on a
line by themselves.

The behavior here differs from other common MTAs.  For example,
postfix's sendmail command documents the -i option very similarly,
but since it does not ordinarily carry out the dot-stuffing behavior
when invoked from the commandline (only when accepting a message
via SMTP), the description of -i in the manpage is accurate for
postfix.

Sendmail's man page should be updated to accurately describe the
effect of the -i option.  Upstream doesn't seem to have a publicly
accessible bug tracker, so I was hoping you could pursue the
issue with them as it exists in the upstream source.


Version-Release number of selected component (if applicable):

8.14.4-4.fc13
(and presumably all other current version-releases)


How reproducible:

100%


Steps to Reproduce:

1.  Create a text file /tmp/testmail.txt with the following contents:

==============================================
From: Your Name <youraddress>
To: Your Name <youraddress>
Subject: Sendmail dot test

This is a test
.. This line has two dots
... This line has three dots
This is the end of the test
==============================================

2.  cat /tmp/testmail.txt | /usr/sbin/sendmail -t -f youraddress

3.  cat /tmp/testmail.txt | /usr/sbin/sendmail -i -t -f youraddress


Actual results:

The mail received from step 2 above has one dot on the line that
had two dots in the text file and two dots on the line that had
three dots in the text file.

The mail received from step 3 above looks exactly like the text
file.


Expected results:

With most MTAs, you would expect both received messages to look
the same as the text file.  However, given what sendmail does with
the message when invoked as in step 2, according to the man page's
description of the -i flag, the message received when it is invoked
as in step 3 should look the same as the one from step 2 (it should
have one dot removed from each line) because the -i documentation
incorrectly claims that it only affects lines that contain dots
alone by themselves.

The man page should be updated to correctly describe the behavior,
perhaps along the lines of

"-i  Do not strip a leading dot from lines in incoming messages, and
do not treat a dot on a line by itself as the end of an incoming
message.  This should be set if you are reading data from a file"