[mailto:owner-ietf-smtp(_at_)mail(_dot_)imc(_dot_)org] On Behalf Of Peter J.
Sent: Tuesday, August 23, 2011 2:09 AM
To: SMTP Interest Group
Subject: Re: productivity?
BTW, I've never observed a sendmail to keep open an idle connection for
that long. Maybe that limit doesn't apply to "normal" connections
(according to the document you cited it only applies to "IPC
connections") or maybe that limit is only reached under rare
circumstances. I just tested an out-of-the-box sendmail-8.13 (with
ConnectionCacheTimeout=5m) and it sends the QUIT immediately after the
250 response to DATA.
Sendmail's connection cache only applies during queue runs. The entire
connection cache is capped at a certain number of connections, which is where
the LRU comes in; when it needs to make a connection to a server that is not
currently cached, the most idle one is closed down and the new one is
established. When the queue run completes, any cached connections remaining
are also closed.
So really you would only ever see a cached connection stay open that long if,
for a connection cache of size N, it had to connect to some small server, but
the remaining (rather large) queue went only to a specific N-1 other servers,
and so the first one sat idle for that long. Seems rather like a corner case.
 I couldn't find out what that is - the old sendmail paper from the
1980's uses the term as a synonym for "socket connection", but I'm not
sure if that's current usage.
"IPC connections" refers to anything that talks SMTP, in that context. The
other mailers (in sendmail parlance) are pipes to local programs.