For example, after that "normalization", quoted printable encoded
headers (as described at rfc 2047) longer than 76 characters are
splitted in two different ill-formed headers because the soft LF line
break becomes a "hard" CRLF one. This is clearly wrong.

rfc 2060 specifically allows CR and LF inside literals:

"""
A literal is a sequence of zero or more octets (including CR and LF),
prefix-quoted with an octet count in the form of an open brace ("{"),
the number of octets, close brace ("}"), and CRLF.
"""

It looks like the IMAP4.append method is responsible for the CRLF
substitution (trunk/Lib/imaplib.py).

# defined near top of module:
MapCRLF = re.compile(r'\r\n|\r|\n')

# in append method:
self.literal = MapCRLF.sub(CRLF, message)

I'll work on a test for it this evening.

-Ron

Module imaplib has pretty sparse test code.  There is only 1 test case,
for imaplib.Time2Internaldate.
trunk/Lib/test/test_imaplib.py

The attached patch tests for LF, CR preservation with regard to the
IMAP4.append method, but more testing is necessary to make sure LF, CR
are preserved in IMAP literals throughout imaplib and that the
respective fix doesn't break anything.  Mock IMAP interaction may be the
way to go for better test cases.  Is anyone working on expanding imaplib
test coverage?

-Ron

Am quite new here; just searching hard to contribute, would like to
patch this, if I can go ahead.

Don't we need to patch the original imaplib code also?

just remove the line:
self.literal = MapCRLF.sub(CRLF, message)
and have:
self.literal = message
Or am I missing something completely?