Berkeley and SRI mailers

[utzoo!decvax!ucbvax!unix-wizards]

>From Nowicki at PARC-MAXC Tue Sep 15 18:26:40 1981
 A while ago we were debugging a new mailer from the Stanford Ethernet to the
Arpanet. Mail to Berkeley often got delivered many times.  We eventually
switched to having it relayed through a real Arpanet machine running Tenex,
and seems to work better.  This is one of those situations where neither side is
really at fault, but the protocol is so complicated that no one could possibly
implement all of it.  Thus, if both side do not implement certain features in
combination with other features and sequences of events (like the phase of the
moon) strange things happen.

Yet another argument for simpler protocols!

	-- Bill