'Throughput' should be for the benefit of the receiver, not the sender.
It's inconsiderate to send using as many threads as possible to a single
receiver.
Oh, please, not this old wheeze again. What's really inconsiderate is
to demand that everyone in the world artifically delay their mail
because you don't know how to manage your mail server.
If I have mail for some of your users, I expect that they will have at
least some interest in reading it, so the sensible and considerate
thing to do is to deliver it promptly so they can do so. I have no
way of knowing how your mail system works, so the way to do that is to
connect to it and send the mail.
For reasons that should be obvious to anyone who understands how TCP
works, on the modern Internet, the round trip delay time for each SMTP
command is typically longer than the time it sends to send the
traffic, and mail sessions are so short that no single session will
grow to fills more than a fraction of a typical pipe, so it is a lot
faster to send mail in parallel rather than single threading. (The
bandwidth is trivial, the total amount used by all mail including all
the spam is barely a blip compared to video streaming and p2p.)
So to deliver that mail, I open some mail connections and send the
mail. If for whatever reason your mail system can't or won't accept a
mail session, reject or soft fail it, and any normal SMTP sender will
back off. My mail software can deal with rejection, it's software,
not a person.
On my mail system there is a global SMTP session budget. It accepts
connections whever it's not at the limit, rejects them when it's over,
without worrying about which senders are using which sessions. That
works great, sheds load quite nicely.
I realize that in some circles full speed mail delivery is considered
anti-social, so sending mail software has developed elaborate
heuristics to try to game the receivers and ramp up to whatever number
of threads won't quite get them throttled. But that's religion, not
technology.
R's,
John
_______________________________________________
ietf-smtp mailing list
ietf-smtp(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf-smtp