masqmail

annotate docs/uucp-setup @ 323:29de6a1c4538

Fixed an important bug with folded headers! g_strconcat() returns a *copy* of the string, but hdr->value still pointed to the old header (which probably was a memory leak, too). If the folded part had been quite small it was likely that the new string was at the same position as the old one, thus making everything go well. But if pretty long headers were folded several times it was likely that the new string was allocated somewhere else in memory, thus breaking things. In result mails to lots of recipients (folded header) were frequently only sent to the ones in the first line. Sorry for the inconvenience.
author meillo@marmaro.de
date Fri, 03 Jun 2011 09:47:27 +0200
parents 39014fc31dbe
children
rev   line source
meillo@88 1 This document includes UUCP related information
meillo@88 2
meillo@88 3 Note: My knowledge of UUCP setups is very poor. I hope that the
meillo@88 4 provided information is correct. Improvements to this document
meillo@88 5 are very welcome. --meillo
meillo@88 6
meillo@88 7 UUCP setups call the MTA as `rmail'. Until version 0.2.23 masqmail
meillo@88 8 could be called with this name. It switched to read-message-from-stdin
meillo@88 9 mode then. AFAIK this is not enough to support UUCP, at least not at
meillo@88 10 the level that is presumed by UUCP software. It seems as if at least
meillo@88 11 the first input line should be handled special as it includes the
meillo@90 12 envelope recipient. (Correct me if I'm wrong.) I discovered that exim
meillo@90 13 has rmail support like this. That's probably the reason why it was
meillo@90 14 the same in masqmail.
meillo@88 15
meillo@88 16 A better, through still basic approach, was introduced with 0.2.24:
meillo@88 17 misc/rmail is a small shell script (taken from postfix), which calls
meillo@88 18 masqmail with appropriate options. Copy the script into your path and
meillo@88 19 ensure that the included sendmail variable points to the masqmail
meillo@88 20 executable.
meillo@88 21
meillo@88 22 A more sophisticated rmail implementation seems to be available from
meillo@88 23 sendmail. I don't know details about it and whether it is needed. The
meillo@88 24 difference of sendmail's rmail implementation could be related to
meillo@88 25 address rewriting (user@example.org <-> org!example!user). But I
meillo@88 26 don't know details -- if you do, please let me know.
meillo@88 27
meillo@88 28
meillo@88 29 UUCP makes use of the -f (set return path address, i.e. from whom the
meillo@88 30 mail is) option of masqmail which is only permitted for user root,
meillo@88 31 the trusted user (usually `mail'), and the trusted group (often group
meillo@88 32 `mail'). UUCP, however, usually runs as user and group `uucp'.
meillo@88 33
meillo@88 34 Masqmail currently supports only one trusted group and it is planned
meillo@88 35 to remain so for simplicity reasons. (If you have good arguments on
meillo@88 36 the case, try to convince me of the opposite.) Therefore the solution
meillo@88 37 for masqmail is to add the user `uucp' to the trusted group (often
meillo@88 38 group `mail'):
meillo@88 39
meillo@88 40 usermod -G mail -a uucp
meillo@88 41
meillo@88 42 This is not the perfect solution but an acceptable trade-off.
meillo@88 43
meillo@88 44
meillo@88 45 If one really needs to enable user `uucp' to set -f but can not add
meillo@88 46 it to the trusted group, see the comment in is_privileged_user() in
meillo@88 47 permissions.c. It shows a hack which allows to trust another group,
meillo@88 48 for instance the group `uucp'.
meillo@88 49
meillo@88 50
meillo@88 51 See [1] for reasons why -f is important and needed.
meillo@88 52
meillo@88 53 [1] http://bugs.hylafax.org/show_bug.cgi?id=842
meillo@88 54
meillo@88 55
meillo@88 56 meillo