OSEC

Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com
 
From: Amos Gouaux (+archive.postfix-usersutdallas.edu)
Date: Fri Feb 16 2001 - 09:03:51 CST

  • Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

    >>>>> On Fri, 16 Feb 2001 09:07:29 +0100 (MET),
    >>>>> Alexander Keller <kelleristac.de> (ak) writes:

    ak> I do not know any details about the lmtp protocol, but as I understand the
    ak> above conversation, the cyrus lmtp server is advertising some of it's
    ak> features, but does not tell the postfix lmtp client to authenticate. Since
    ak> I do not know if this is necessary, may be I have a cyrus problem
    ak> here?? Where should postfix know, that it should authenticate??

    This is in fact the problem: the Cyrus lmtpd is not advertising AUTH
    capability, and so Postfix doesn't attempt it. In my tinkering last
    night before I started falling asleep I noticed that--at least for
    me--if I had

      lmtp cmd="lmtpd" listen="[127.0.0.1]:lmtp" prefork=1

    The AUTH capability was not advertised, even though AUTH is required
    to connect to the service. On the other hand, the setting

      lmtp cmd="lmtpd -a" listen="[127.0.0.1]:lmtp" prefork=1

    did display the AUTH capability even though the "-a" pre-authed the
    service so that the LMTP-AUTH communication wasn't required.

    What's odd is that the Cyrus code in imap/lmtpengine.c looks
    reasonable. Then I feel asleep.

    Oh, and another thing I seemed to stumble upon is that, as some
    folks have pointed out on info-cyrus, it would appear that the
    Postfix lmtp client is serializing the recipients.

    With Postfix snapshot-20001030, if there are multiple recipients in
    a message, and they all have the same destination, they will all be
    listed in "RCPT TO" lines during a single LMTP transaction. With
    Cyrus this is important because this translates into a (possible)
    single-instance message store delivery.

    However, with snapshot-20010128, it would appear that all these
    recipients are serialized into separate LMTP transactions, which can
    seriously degrade performance. I tried with/without compiling with
    the SASL libs. I also tried the lmtp:unix: and the lmtp:host
    delivery methods, trying both mailbox_transport and a transport map.

    I'll try to do some more fiddling later today.

    -- 
    Amos