Hey,
I've been distracted from developing a qmail test-bed with the following
Sendmail problem, and I'm concerned about the same problem in qmail.
We're using the popauth hack (for roaming users) and the DUL-RBL. This
works great, unless a popauth-OK'ed address is in the DUL. The DUL is
checked first. Since checks end with a reject on matches to the rbl, the OK
in popauth.db is never seen, and all outgoing mail from the user bounces,
even though they've authenticated via POP.
I understand that things flow in the opposite direction here, but it seems
like the problem remains-- at least according to my rudimentary knowledge of
qmail. First, tcpserver checks tcp.smtp.cdb and passes the connection with
x.x.x.x:allow,RELAYCLIENT="". Later in the pipe we encounter instances of
rblsmtpd. $RBLSMTPD is not set, so a lookup is done, and matched against
the DUL, and the mail is rejected. rblsmtpd doesn't care about the status
of $RELAYCLIENT, and the message never reaches qmail-smtpd, because the pipe
has already closed.
If I were using a simple /etc/tcp.smtp file, I'd gather the answer would be
to modify lines to read x.x.x.:allow,RELAYCLIENT="",RBLSMTPD="". Perhaps
this is a solution for customers using our dial-ups. But I'm concerned with
roaming users. I'm using Vpopmail with --enable-roaming-users, and this
adds its own lines to the cdb.
Would I need to modify the Vpopmail source to add the RBLSMTPD=""
assignment? Am I imagining bogymen and there's no problem? Or is there a
problem, but with a simple solution I've missed?
Thanks,
Mark Tippetts
