[2310] in Commercialization & Privatization of the Internet
daemon@ATHENA.MIT.EDU (postman@nisc.psi.net)
Thu Feb 6 10:16:18 1992
To: marc@Athena.MIT.EDU
Cc: com-priv-redist@MIT.EDU
Date: Thu, 06 Feb 92 10:15:28 -0500
From: postman@nisc.psi.net
com-priv-redist@mit.edu is on the list. I suggest trying send to the list
locally. I'm cc'ing this message to it.
Brad
------- Forwarded Message
Return-Path: marc@Athena.MIT.EDU
Received: from psi.com by nisc.psi.net (5.61/2.1-PSINet Operations )
id AA03500; Wed, 5 Feb 92 19:01:07 -0500
Received: from ATHENA.MIT.EDU by psi.com (5.61/2.1-PSI/PSINet)
id AA20139; Wed, 5 Feb 92 19:01:00 -0500
Received: from HODGE.MIT.EDU by Athena.MIT.EDU with SMTP
id AA03397; Wed, 5 Feb 92 19:00:55 EST
Received: by hodge (5.57/4.7) id AA17242; Wed, 5 Feb 92 19:00:52 -0500
Message-Id: <9202060000.AA17242@hodge>
To: com-priv-request@psi.com
Cc: postmaster@psi.com
Subject: [Pushpendra Mohta: Re: SLIPing Questions and Answers (fwd) Here it is]
Reply-To: Marc Horowitz <marc@MIT.EDU>
X-Usmail: Marc Horowitz, 3 Ames Street, Cambridge, MA 02139
X-Phone: (617)225-6448
Date: Wed, 05 Feb 92 19:00:49 EST
>From: Marc Horowitz <marc@Athena.MIT.EDU>
I complained about a week ago that some postings of mine to com-priv
weren't making it back to me. Well, I have some concrete evidence now
:-)
I sent a posting to com-priv@psi.com. I know it was dist'ed to at
least some people, because people have responded. But it was never
sent to com-priv-redist@athena.mit.edu (possibly @mit.edu, they run
the same aliases file), our local exploder. I had someone forward me
back my posting, which is included below. It's got a message-id, so
hopefully you can track it down in your logs. Please respond if you
can figure out what is wrong. Thanks.
Marc
- ------- Forwarded Message
>From: Pushpendra Mohta <pushp@CERF.NET>
Subject: Re: SLIPing Questions and Answers (fwd) Here it is
To: marc@MIT.EDU
Date: Sun, 2 Feb 92 14:44:04 PST
Marc Horowitz writes:
>From com-priv4-forw@psi.com Sun Feb 2 00:28:47 1992
Message-Id: <9202020820.AA09981@portnoy.MIT.EDU>
To: bzs@world.std.com (Barry Shein)
Cc: com-priv@psi.com, cook@tmn.com
Subject: Re: SLIPing Questions and Answers
In-Reply-To: [2236] in Commercialization & Privatization of the Internet
Reply-To: Marc Horowitz <marc@MIT.EDU>
X-Usmail: Marc Horowitz, 3 Ames Street, Cambridge, MA 02139
X-Phone: (617)225-6448
Date: Sun, 02 Feb 92 03:20:55 EST
>From: Marc Horowitz <marc@Athena.MIT.EDU>
>> I proposed a model which minimizes administration and could support
>> around 32 simultaneous sessions (depending on what's actually being
>> done) for a relatively low price, by using terminal servers across
>> 56kb lines. It could be easily duplicated for N*32 seats. But it
>> didn't involve any SLIP. Local hardware and administration is
>> minimized.
For real interactive use, I don't think the terminal server model is
right. I don't think having a local news/mail/foo server is right
either. The interface we want is something which gives you the ease
of running remotely, and the speed/power of running locally. In other
words, you need some very clever user interfaces, which I think would
run just as well over an IP link (SLIP/PPP/whatever). A serial link
running one application at a time doesn't cut it. It's natural for a
user to be reading news, and say, "I want the file mentioned in that
post", and start the transfer. Maybe it's slow, but if it runs while
you're doing something else, you don't care as much.
I think the problem is that Barry is assuming that when someone says
"TCP/IP" they mean NNTP, SMTP, FTP, and all those other protocols
which, when they break, require someone who knows what they are doing
to fix. I think the effort belongs in writing clients which optimize
the use of the link between the user and the server. For some
applications, like ftp, there are already clients for PC's which do a
good job. For others, say, news, you might want a news reader which
runs partially on some central server, and partially on the local
host. I get to see headers, my kill files are run near the articles,
not over some slow link, and only the articles I actually read are
pushed over (even prefetched), but they're cached, so I can go back
and forth within or between posts without a real delay. Do it "right"
and those of us who are sitting on the same wire as our news servers
can use the same application. That's why I think IP is the right
protocol. Having one protocol is a *big* win. The problem is that
most of our data protocols were designed with big (ethernet) networks
in mind. We need new data protocol which work well over slower links,
not new network protocols.
Now, I'm just thinking in real time here. I'd probably change some of
these details twelve times if I were to go write something, but the
point is, that SLIP is fine. PPP is fine. We need plug'n'play
applications *designed for slower links*, and that is where
development should go. (Someone will probably respond that these
exist. Great. Make sure everyone knows about them. If they're
really good, I'll want to use them when I'm on a fast net, too.)
(Gee. Maybe someone should offer to put free lines into developer's
homes, forcing them to design useful protocols :-)
Marc
- - --pushpendra
Pushpendra Mohta pushp@cerf.net +1 619 534 5056
CERFNet
- ------- End of Forwarded Message
------- End of Forwarded Message
