Received: from localhost by CS.UTK.EDU with SMTP (cf v2.9s-UTK) id LAA14775; Wed, 6 Mar 1996 11:38:25 -0500 Received: by CS.UTK.EDU (bulk_mailer v1.4); Wed, 6 Mar 1996 11:37:15 -0500 Received: from koobera.math.uic.edu by CS.UTK.EDU with SMTP (cf v2.9s-UTK) id LAA14718; Wed, 6 Mar 1996 11:37:12 -0500 Received: (qmail-queue invoked by uid 666); 6 Mar 1996 16:39:02 GMT Date: 6 Mar 1996 16:39:01 GMT Message-ID: <19960306163901.14313.qmail@koobera.math.uic.edu> From: djb@koobera.math.uic.edu (D. J. Bernstein) To: drums@cs.utk.edu Subject: Re: re PIPELINING > My experience contradicts this. Really? What percentage benefit could your host possibly get from multiple RCPTs? For outgoing messages this is very easy to measure: for each message, count the number of recipients and the number of hosts; scale by message size if you want; add up the results for all messages. Note that the situation is different for non-Internet-connected hosts. For example, hosts behind a firewall typically have to send all messages through a gateway. They should use a protocol designed for that---like the BITNET protocol. :-) > We run a mailing list with 5000 recipients. [ best performance: 20 domains per envelope; 2-hour delivery overall ] qmail sustains performance of <1 second per SMTP connection with its default concurrency limit of 20; that's already <2 hours for your 5000 recipients. If most of your recipients are in the U.S. you'll do even better. You can, of course, use a much higher concurrency limit. ---Dan