[10801] in Commercialization & Privatization of the Internet
re: How Long to a Multimedia Internet
daemon@ATHENA.MIT.EDU (Craig Partridge)
Thu Mar 10 03:19:21 1994
To: com-priv@psi.com
From: Craig Partridge <craig@aland.bbn.com>
Date: Wed, 09 Mar 94 17:02:41 -0800
>how long will it be until mosiac style internet interfaces are the
>most common form of internet access?
I don't know about mosaic, but I can talk a little bit about other forms of
multimedia, namely voice and video transmission over the Internet.
If you want to do one way voice (e.g., lectures) you can do it now.
The limitation is largely how much buffering you have at the receiving
end to resequence the audio samples. The IETF is multicast routinely now.
For different reasons, two way voice and one and two way production quality
video need IP to be enhanced a tiny bit. (Briefly, routers have to recognize
certain IP datagrams as special, either by tagging the datagrams or putting
special pattern matchers in the routers and then throwing those datagrams
in different queues which get different service. Routers also have to be
able to say "no, I can't take another high-bandwidth application" or "no,
I can't promise you that low a one-way delay" to applications). Much of
the important theoretical work has been done, and experimental systems that
illustrate some of what can be done are being tested by researchers.
Note that if this sounds like making IP more like the telephone system,
you're mostly wrong. The algorithms are different and so is the philosophy.
The goal of most of the proposed IP systems is to allow the system to recover
gracefully if a router fails (your phone call doesn't disconnect when a switch
fails), and information in routers is soft (it can be discarded or lost and
restored later). Finally, rather than focussing on supporting 64Kb/s pipes
(or multiples thereof), the idea is to support requests for arbitrary
bandwidths and probably a range of delays. But the result is that all
sorts of things like video on demand, voice and video calls from your desktop,
etc., will be available over the Internet.
Of course, this isn't going to come immediately. The working groups
will kick off their first meetings the end of this month (at IETF in Seattle).
And their timetables call for starting to deliver proposed standards in late
1995. But deployment in the Internet can be quite brisk, so you might
conceivably see people other than researchers using the stuff sometime in 1996.
Craig
(Chair of one of the soon-to-be-announced IETF WGs in this area)
