Newsgroups: comp.protocols.tcp-ip
Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!malgudi!osc.edu!karl.kleinpaste
From: karl.kleinpaste@osc.edu
Subject: Re: compuserve
Message-ID: <1991May15.143317.4297@oar.net>
Sender: news@oar.net
Nntp-Posting-Host: ashley.osc.edu
Organization: Viento Gigabit Testbed, Ohio Supercomputer Center
References: <1991May15.054958.16231@Spies.COM>
Date: Wed, 15 May 1991 15:32:29 GMT
Lines: 52

joshua@Spies.COM writes:
   |>The only connection to CompuServe is for email passage.

   And their mail is semi-frequently hosed. For the amount of money
   they charge their subscribers, it does not seem to me that said
   subscribers get incredibly good service. 

[adjusting mailer-daemon cap.  (Yes, I really have such a thing, a
weird blue-and-purple hat with horns on it that Elizabeth Zwicky
crocheted for me a couple of years back.  It makes me fit the part.)]

Consider that the situation is the following...

1. The link to CServe is a 9600bps modem and a poor protocol.
2. The link is passing almost 2 orders of magnitude more traffic per
day than was ever anticipated.  That'll teach us...
3. The queue which builds up during the nominal working day now
finally manages to clear around 4am EST the next morning...just in
time for the Europeans using it to start generating the next day's
blortful.
4. There are CServe users who are requesting large(!) things from
places like bitftp@pucc.princeton.edu.  That MBASes like bitftp are
tolerated at all in non-load-sensitive incantations is a mystery to me.
5. When the backlog gets too great, the load on the MX hosts too high,
and the discs there too full, we have to shut down the mailers while
we try to get things to clear somehow.  (We do _not_ shut down the
mailers for "days at a time," as a rumor I heard described it.)
6. When things go completely insane, as they did about 2 weeks ago, we
ultimately boot up SneakerNet and deliver, oh, about 8000 messages to
CServe via magtape.  Probably closer to 12000 this last time, actually.
7. The support for the link on the Internet side is strictly volunteer.
8. CServe is taking steps to move the gateway entirely in-house, so as
not to depend on volunteer assistance, and to gain the advantages of
higher link speeds and possibly faster protocols.  I've been
contracted to help with the software conversion and assist CServe
people in some of the questions of maintaining an IP connection.  They
have chosen a commercial IP provider, and the lines are (or should be)
ordered.

In the meantime, the current house of toothpicks is occasionally blown
over and we have to go whittle a new set to build it up again.  Just a
little patience, we'll get there.  And apologies for the occasionally
degraded/delayed service in the meantime.

--karl kleinpaste
Personification of the Internet (Light) Side
of the Schizoid CompuServe Mailer Daemon
(and the CompuServe (Dark) Side is itself
 schizoid-split as well, imagine that)

PS- No, emphatically, general IP (telnet) access into CompuServe is
_not_ part of the game plan.