[8424] in Commercialization & Privatization of the Internet
Re: Voice telephony on the Internet?
daemon@ATHENA.MIT.EDU (Craig Partridge)
Thu Nov 18 13:40:10 1993
To: "Bill St. Arnaud" <wcsv2k@ccs.carleton.ca>
Cc: com-priv@psi.com
In-Reply-To: Your message of Thu, 18 Nov 93 13:10:06 -0500.
From: Craig Partridge <craig@aland.bbn.com>
Date: Thu, 18 Nov 93 10:38:31 -0800
Is anybody looking to see if any of the proposed IPng's can carry
constant bit rate traffic? How about the router's, etc etc? What are
the issues in migrating from a store and forward to real time or
hybird architecture?
Bill:
The answer is an emphatic yes, folks are doing voice (and video)
conferencing over the Internet using UDP/IP and TCP/IP and have been doing
so for a few years. This conferencing is both small scale (between two persons)
and wide scale (delivery to hundreds of people simultaneously). For instance,
the Internet Engineering Task Force (IETF), transmits its quarterly meetings
(both voice and video) all over the Internet for folks who can't come to the
meetings in person. The last meeting (early this month) had 600+ listeners
from all over the globe.
I believe the most commonly used software is the VAT software, which
can be anonymously FTPed from ftp.ee.lbl.gov but other software exists.
Some of the this software is compatible with VAT (e.g., NEVOT), while other
software implements other experimental protocols (e.g., DEC is doing some
interesting work with video conferencing).
There's also been a lot of good theoretical work on algorithms for
packet scheduling. Rather than cite it all, I'll point you at chapters
11-13 of my new book, "Gigabit Networking" (Addison-Wesley, ISBN
0-201-56333-9), since I wrote those chapters to answer questions like
these.
Personally, I believe the key results of this research to date are:
(1) applications like voice can adapt surprisingly well to variations in
delay; (2) multicasting services in the network are tremendously important
for effective conferencing; and (3) performance would improve if the network
was able to at least partially isolate real-time traffic from regular
traffic, but the isolation probably does not need to be very strict.
Craig Partridge
