Problem with virtualdomains and VERPs (with patch)

[Patrick J. LoPresti]

qmail-send handles virtual domains (by prepending the virtual domain
prefix) before it creates a Variable Envelope Return Path.

This creates problems for me on my master mail server.  For example,
if I am the user "[EMAIL PROTECTED]", and I address a message to
"[EMAIL PROTECTED]", and I request a VERP (say, by setting the "r" flag in
QMAILINJECT), and control/virtualdomains has the following line:

foo.com:blah

...then the VERP reads "sender-blah-recip=foo.com" instead of the
correct "sender-recip=foo.com".

Incidentally, I noticed this because ezmlm is not removing dead
accounts from my mailing lists because the VERPs are wrong.

The appended patch seems to fix the problem.

 - Pat

diff -u -r1.2 qmail-send.c
--- qmail-send.c2001/04/23 15:40:29 1.2
+++ qmail-send.c2001/04/30 17:00:25
@@ -171,6 +171,7 @@
  int j;
  int k;
 
+ recip = stripvdomprepend(recip);
  i = str_len(sender);
  if (i >= 4)
if (str_equal(sender + i - 4,"-@[]"))

Qmail, double-bounces, and RFC2821

[Patrick J. LoPresti]

(I searched the mail archives briefly but did not see any discussion
of this.  My apologies if I missed it.)

This is the new RFC which supersedes RFC821 as the SMTP specification:

  http://www.faqs.org/rfcs/rfc2821.html

The grammar in sections 4.1.2 and 4.1.3 appears not to permit [] as
the domain portion of a mailbox in an address.  In particular, the
address "#@[]", which Qmail uses as the envelope sender for
double-bounces, is not syntactically valid according to this grammar.

Feel free to double-check me on this, as I would be happy to be wrong.

Comments?  Thoughts?

 - Pat

P.S.  Incidentally, the Lotus Domino SMTP server already rejects Qmail
double-bounce messages as syntactically invalid.  Until a few days
ago, I could claim that Domino was in violation of the relevant RFC
(821).  But by this new RFC, Domino is right and Qmail is wrong.  And
now that RFC2821 has been released, I fear that other MTAs may start
rejecting these messages too.

Bounces should have a Message-ID

[Patrick J. LoPresti]

I know, I know, not all software generates a Message-ID.  But friendly
software does, and it is nice for the MTA to be friendly.

So consider this an enhancement request for Qmail.

 - Pat

Why not permanent failure code for bare LF?

[Patrick J. LoPresti]

I understand why qmail rejects messages containing a bare LF.

My question is, why does it give SMTP result code 451 (indicating
temporary failure) instead of a code to indicate permanent failure?
Sending that same message will fail every time, will it not?

I am just curious about the rationale.

Thanks!

 - Pat