Dave CROCKER wrote:
Here's the essential bit: Occasional connectivity fits into the
Internet mail model only as a last-hop, receive-side issue. There must
be an always-connected server working on behalf of the
sometimes-connected receiver and holding mail for it. Hence, use of
ODMR is a 'local' matter. Do not expect random, remote servers to
participate.
The HUB!
[1] http://isoc.org/wp/newsletter/?p=1098
DISCO was dead by this point! <g>
[2] http://bbiw.net/recent.html#PostelCSNet
[3] It was interesting presenting the SMTP "working group" with the
challenge of this operational style, since it was entirely foreign to
the existing world of the Arpanet. Folks were not all that receptive to
it until Vint Cerf, then at Arpa and always looking for ways to justify
funding worthy projects, suddenly said "submarines!" At that point,
everybody grok'd the operational model that leaves availability for
delivery as a choice of the receiver.
Very interesting David. Fidonet had a Policy4 membership ZMH (Zone
Mail Hour) requirement to allow routes via your hub and no humans were
allowed to tie the line during ZMH.
http://www.fidonet.us/policy4.html
I guess the 821 question I have is, Given Cerf's direction, I don't
see any description or is hidden somewhere in 821 about connection
errors and retries recommendations. Was that because not enough
intelligence yet or you guys just had a "Card Blanc" with full time
reliable connectivity (I suspect delays was par)?