[ih] Nit-picking an origin story

[Craig Partridge]

On Mon, Aug 18, 2025 at 8:36 AM Dave Crocker via Internet-history <
internet-history at elists.isoc.org> wrote:

> On 8/18/2025 7:00 AM, John Day via Internet-history wrote:
> > This last one I think doesn’t get enough credit. It is a very small
> thing, but I think was a major contribution to the success of the ARPANET.
> It would have worked at 2.4 or 9.6, but been so glacially slow as to have
> been considered not successful.
>
> For the general, 'shared resources' model, sure.  And from today's
> perspective, OMG you bet!
>
> Howerver...
>
> CSNet gatewayed Arpanet/Internet mail and initially it was dial-up
> 300baud or 1200 baud.  It it was effective.  And 2.4K would have been
> luxurious.
>
>
As a former CSNET techie (person who had to oversee the CSNET mail
gateway)...

It worked just fine to circulate email among CSNET sites.  But a lot of
CSNET email went out to folks on ARPANET.  And the CSNET-RELAY would have
swiftly
overflowed its disks if there wasn't a 56Kbps pipe to the ARPANET sites
to drain that email swiftly.  The SMTP channel crashing was one of the top
reasons I'd get paged in the middle
of the night (because disks were at risk of overflow of backed up email).
Basically another one of those 80/20 or 90/10 rules -- ARPANET was sinking
80% of the CSNET
traffic and needed a bigger channel.

Craig

PS: Details for those unfamiliar.  CSNET ran as a dialup service.  It had a
bank of modems and a favorable phone contract and would call each CSNET site
twice during the night (night because, in the US at the time, phone rates
dropped sharply between 5PM and 8AM).  The first sweep through the sites
typically
involved picking up accumulated email from each site (pickup was the first
part of the call), for forwarding both within CSNET and to ARPANET (and
later UUNET)
and then delivering accumulated email from ARPANET (and as wegot deeper in
the site list, from sites already called). The second sweep was typically
shorter
and focused on pushing out email from other CSNET sites that had been
picked up earlier in the night.  The result was everyone got email within
24 hours (in urgent cases,
sites could call a CSNET techie and ask for a daytime call. They got
charged more for it, but it meant there was a way to send urgent
emails/responses the same day).

Things got a little wilder when the UUNET gateway (seismo) came up on
ARPANET, as seismo was doing similar things, but due to the hop-by-hop
dynamics of
UUNET, traffic from UUNET to CSNET tended to heat up some hours after 5pm
(c. 10pm if memory serves), aka deep into CSNET's first dialing sweep.
This meant CSNET's second sweep got longer and we sometimes had to deal
with deliveries running past 8AM and paying higher phone rates.

PPS: File under odd thoughts.  Both CSNET-RELAY and SEISMO were in the
Eastern time zone, so the dialup relayed email starting flowing into
ARPANET (and to CSNET sites)
at 5PM Eastern, which was 2PM in California.  So, while East Coast folks
went home for dinner (and remember, many of us didn't have easy email
access at home at the time) as the
relayed email began flowing, the West Coast folks would be reading and
replying to relayed email in time for CSNET/SEISMO to forward the reply
that same night.  I periodically
wonder what the impact was of that response-time advantage....


-- 
*****
Craig Partridge's email account for professional society activities and
mailing lists.