[11307] in Commercialization & Privatization of the Internet
Re: How Long to a Multimedia Internet?
daemon@ATHENA.MIT.EDU (Christopher Davis)
Mon Mar 28 20:04:16 1994
Date: Mon, 28 Mar 1994 11:36:46 -0500
From: Christopher Davis <ckd@kei.com>
To: tenney@netcom.com (Glenn S. Tenney)
Cc: com-priv@lists.psi.com
In-Reply-To: <199403262312.PAA04054@netcom9.netcom.com>
GST> == Glenn S Tenney <tenney@netcom.com>
GST> On that same wavelength... The Xmodem protocol is fine for very
GST> narrow and long pipes too, but I wouldn't want to use it for ftp and
GST> telnet and www and mosaic-like applications...
HTTP is, for all intents and purposes, just a TCP stream with a little bit
of format labeling and negotiation at the beginning. TCP is pretty well
tuned to run on links from 9600bps on up past T1 to FDDI, at least IMHO.
XMODEM is *not* fine for "very narrow and long pipes too" as it lacks
sliding windows and is lock-stepped at 128 byte blocks (1K if you're using
that extension).
Jim Thompson's point, that the data being served--not the protocol--is the
problem is correct. No matter how wonderful the sliding windows, there's
no way to feed an 800K ulaw file down a V.32bis modem in a reasonable
amount of time. However, a 20K HTML file will take very little time (and
probably compress pretty well over V.42bis also).
http://www.kei.com/internet-tour.html is less than 8K, yet it includes a
fairly rich set of links, if I do say so myself. It's less than half the
size of my USENIX FaceSaver picture in GIF format. Yet both of those, and
many much larger items, are served via HTTP.
