Q: Sendmail -- working around overly long greet pauses

[Philip Prindeville]

I'm trying to deliver email to some sites that seem to have
obnoxiously long greet pause configured (yes, it's a
Chinese site, and yes, it's their contact address for reporting
abuse... guess they aren't about to make it easy to do):

# date ; sendmail -q -v ; date
Tue Feb 20 12:41:19 MST 2007

Running /var/spool/mqueue/l1J1BZTd015129 (sequence 1 of 4)
<anti-spam at ns.chinanet.cn.net>... Connecting to ns.chinanet.cn.net. via esmtp...<anti-spam at ns.chinanet.cn.net>... Deferred: Connection timed out with ns.chinanet.cn.net.

Running /var/spool/mqueue/l1HKVCiu004463 (sequence 2 of 4)
<anti-spam at ns.chinanet.cn.net>... Deferred: Connection timed out with ns.chinanet.cn.net.

Running /var/spool/mqueue/l1HKWqcD004481 (sequence 3 of 4)
<anti-spam at ns.chinanet.cn.net>... Deferred: Connection timed out with ns.chinanet.cn.net.

Running /var/spool/mqueue/l1GKJkvD011374 (sequence 4 of 4)
<anti-spam at ns.chinanet.cn.net>... Deferred: Connection timed out with ns.chinanet.cn.net.
Tue Feb 20 12:44:29 MST 2007
#


I've tried bumping the connect timeout:

define(`confTO_CONNECT', `5m')dnl

(which results in "O Timeout.connect=5m")...  as well as
running sendmail as:

sendmail -q -v -O Timeout.connect=5m -O Timeout.initial=5m

but this doesn't seem to make any difference.  It looks like
the kernel's internal maximum timeout on a connect() is forcing
a shorter interval.

Anyone know what the workaround for this is?  Do I need to force
the kernel's TCP connect() timeout to something larger?

Thanks,

-Philip