[Raw Msg Headers][Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: zmailer / pidentd

On Mon, May 29, 2000 at 03:22:31PM +0200, Remco van de Meent wrote:
> Hi,
> I have a little problem/question regarding the zmailer mail server.
> A popular ident-daemon called 'pidentd' is able to handle multiple
> queries during one connection. This kind of behaviour is also
> documented in RFC1413 as shown below this message.

   The more I look at the thing, the more I am inclined to recommend
   of *not* using ident lookup  (smtpserver's "-a" option) at all.

   This is of same kind of thing as tcp-wrapper; it used to be sensible
   package in the old days, but now tcp-wrapper packs horrendous NIS+ junk
   in without an ability of not needing that "little extra" library set
   for it...

> What happens that upon connection to zmailer server from a host with
> pidentd running, it takes quite some time (>15 minutes in older zmailer
> releases, currently the waiting time is about 10 seconds:
> $ date ; echo quit | socket speer.nl.uu.net 25 ; date
> Mon May 29 15:15:21 CEST 2000
> 220 speer.nl.uu.net ZMailer Server 2.99.52-pre3 #1 ESMTP+IDENT ready at Mon, 29 May 2000 15:15:31 +0200
> 221 2.0.0 speer.nl.uu.net Out
> Mon May 29 15:15:33 CEST 2000
> During this waiting time, the remote zmailer server has en open
> (CONNECTED) connection to the ident/auth port of the local machine
> running pidentd.
> With previous version (2.99.49 for example), the waiting time was that
> long that mail did not get delivered at all because of that waiting
> time (timeouts in the MTA trying to send mail to the zmailer server),
> but luckily this isn't the case anymore.
> I didn't have a very close look on the zmailer code, but does it wait
> for the server to close the connection or so? Maybe it is better if
> zmailer would send a "quit" message or something like that - it would
> be ignored by ident-daemons not knowing about extensions to the basic
> ident protocol I guess.

  Calling  close(sock)  on the ident query client socket is also sensible
  alternate.  It was quite while ago when I poked things to that identuser.c
  client code, and made sure it times out in decent time.

  Looks like it predates even initial import to CVS..

date: 1998/02/10 21:01:46;  author: mea;  state: Exp;  lines: +0 -0
First import of ZMailer to CVS

> It is not something I'd worry about too much, but maybe it is of
> interest to someone.. :)
> Kind regards, 
> Remco.

/Matti Aarnio	<mea@nic.funet.fi>