Received: from localhost by CS.UTK.EDU with SMTP (cf v2.9s-UTK) id MAA18257; Wed, 6 Mar 1996 12:15:45 -0500 Received: by CS.UTK.EDU (bulk_mailer v1.4); Wed, 6 Mar 1996 12:15:07 -0500 Received: from koobera.math.uic.edu by CS.UTK.EDU with SMTP (cf v2.9s-UTK) id MAA18014; Wed, 6 Mar 1996 12:15:05 -0500 Received: (qmail-queue invoked by uid 666); 6 Mar 1996 17:16:57 GMT Date: 6 Mar 1996 17:16:57 GMT Message-ID: <19960306171657.14365.qmail@koobera.math.uic.edu> From: djb@koobera.math.uic.edu (D. J. Bernstein) To: drums@cs.utk.edu Subject: Re: Multiple RCPT-TOs > Conserving bandwidth rather > than latency seems like the right thing to do in the general case, > though not in every case. It's the right thing to do if you're already within a few percent of your maximum network capacity. That isn't the situation for AOL. (Some of you may be amused to note that AOL does about two _hundred_ retries of a message in its first day in the queue.) It isn't the situation for me, and I'd be surprised if it's the situation for any of you. I don't mean to say that bandwidth is irrelevant---but, just as in any other kind of programming, you should _profile_ and spend your time worrying about where most of the bandwidth goes. ---Dan