Path: utzoo!attcan!uunet!csdev!ll1a!spl1!laidbak!att!pacbell!ames!vsi1!apple!desnoyer
From: desnoyer@Apple.COM (Peter Desnoyers)
Newsgroups: comp.sys.next
Subject: Re: Using the NeXT and Ethernet for real-time voice
Message-ID: <8707@spl1.UUCP>
Date: 3 Nov 88 17:35:02 GMT
References: <10865@reed.UUCP> <31814@bbn.COM>
Sender: news@spl1.UUCP
Organization: Apple Computer Inc, Cupertino, CA
Lines: 19

In article <31814@bbn.COM> jr@bbn.com (John Robinson) writes:
>In article <10865@reed.UUCP>, mdr@reed (Mike Rutenberg) writes:
>> [stuff about EtherPhone]

There are two Ethernet research phone systems I know of - Etherphone
at Xerox PARC and a system at Bellcore. If anyone wants references,
email me. Anyway, on the subject of real-time voice and compression:

Speech compression is pretty pointless over a local ethernet unless
you are using huge packets and corresponding delays. 10mS packets
would take 80 bytes at 64kb, compared to well over 40 bytes for TCP/IP
and ethernet per-packet overhead.

Where compression wins big is when you store voice - e.g. annotation,
voice mail, etc. You can do compression over larger chunks, since
delay is not critical, while "bandwidth" - disk space - is much more
expensive.

				Peter Desnoyers