On Wed, 24 Apr 1996, Neal Becker wrote:
> The problem with this theory is that your mailer was not simply
> deferring in sending out the header -- it was listening for messages --
> one could immediately go ahead and enter input and have it accepted.
TCP is a duplex protocol, so naturally it would be listening. The
application won't be reading them, though.
> Also, your mailer disconnects at 5 minutes. If the sender needs
> to be willing to wait 5 minutes for the initial message, your receiver
> should wait longer than 5 minutes for the first message from the sender.
> In fact, you dump the connection at 5 minutes (see below).
I think this is evidence that Nigel is right - it is the *connect*
timeout that is binding. We will have to do some hacking inside the
libident code to do anything about this. However, I am still not sure I
understand this, since Exim doesn't start its listening timer until it
writes the greeting, which it does after attempting to read the ident,
whether or not it manages to get an ident.
> It would appear that when your mail receiver fails in its ident,
> it skips sending the greeting, and goes immediately to listening
> for input without any 220 greeting.
No, that is not the case. The code just doesn't look like that. I have
made a note to do some experiments.
--
Philip Hazel University Computing Service,
ph10@??? New Museums Site, Cambridge CB2 3QG,
P.Hazel@??? England. Phone: +44 1223 334714
