W3C home > Mailing lists > Public > www-talk@w3.org > March to April 1995

Realtime audio on the Web

From: James Gwertzman <gwertzma@das.harvard.edu>
Date: Wed, 12 Apr 1995 19:53:30 -0400
Message-Id: <199504122353.TAA24624@cognac.harvard.edu>
Cc: www-talk@www10.w3.org
I have been researching caching on the Web for a while now, and have
found that using both proxy caches as well as more sophisticated
server-based replication methods, network bandwidth can be reduced by
a factor of 60% or more. A draft of my thesis can be found
at:  http://das-www.harvard.edu/research/vino/push.cache/
or   ftp://steward.harvard.edu/users/gwertzman/thesis.ps.gz
for those who are interested.

I am worried that the real-time audio system will be used for
application that do not actually require real-time audio. There are
advantages to downloading a file first; namely that it is a cacheable
entitiy. A real-time audio stream by nature can not be cached. It may
be argued that this is a security precaution, that music companies do
not want their music to be easily copied, but it seems that it should
be trivial to re-record an audio stream as it is being received and
save it to a file.

It could be argued that network bandwidth is always growing, but it
seems foolish to knowingly waste it. There are several ways that the
Web community is already wasting bandwidth: html pages could easily be
compressed, for example, as part of the protocol, and uncompressed as
they are received. Likewise I am looking forward to certain dynamic
pages being made available as scripts that can be cached and
replicated. It is cheaper for a script to reference a piece of shared
data at the primary host than it is for the browser to retransmit the
entire page again with the updated information in place.

if someone knows more about the realtime standard I welcome their
comments; Maybe I'm wrong and it would be possible to cache these
pages. Certainly the original audio file could be replicated using the
techniques I have developed in my work, although proxy caching would
still be difficult.

James Gwertzman.
gwertzma@das.harvard.edu
Received on Wednesday, 12 April 1995 19:53:45 UTC

This archive was generated by hypermail 2.4.0 : Monday, 20 January 2020 16:08:16 UTC