From: wcs@anchor.ho.att.com (bill.stewart@pleasantonca.ncr.com +1-510-484-6204)
Date: Sun, 8 Jan 95 19:54:12 PST
To: weidai@eskimo.com
Subject: Re: Latency, bandwidth, and anonymity
Message-ID: <9501090353.AA13655@anchor.ho.att.com>
MIME-Version: 1.0
Content-Type: text/plain


My initial reaction to "Anonymous video conferencing" was
"That's when you wear black ski masks and use voice scramblers
and call from video payphones", i.e. not very useful.
("Subcomandata Marcos here...")

On the other hand, Wei Dai's followup message about
> In the longer term, anonymous communication is in danger of being used
> only by fringe groups if it falls too much behind the non-anonymous kind
> in terms of latency and bandwidth (and cost, I guess).  Maybe ONLY drug
> dealers, nuclear terrorists, etc., will use anonymous remailers when full
> sensory virtual interaction is the must popular way for most people to
> communicate and remailers are still the only choice for the
> anonymity-conscious.
puts a different spin on it.  It's a real problem, if not now, then
maybe in 5-10 years.  I realize that those of us in the Phone Company
who have predicted universal Picturephone in the past have been 
over-optimistic :-), but the video compression people and the faster-chip
people keep bringing us closer to having good-quality low-bandwidth video,
and ISDN and fast modems are bringing available loop-end bandwidth up
to the point that reasonably-priced circuits can carry it.  (Long-haul
raw bits have been cheap enough for a while; it's the distribution
and switching technology that have a lot of the cost, and providing
cheap high-bandwidth circuits makes it hard to make money on voice calls.)

The approaches to anonymous video conferencing will depend a bit
on whether the technology takes off on the nets or the phone system,
if those two are still different by then.  It's easier to obscure
the origins of a call on the nets, where users own large parts,
than it is on the phone system, where the Phone Companies own
and operate most of it; the latter environment would require
Phone Remailers, such as PBXs you call into on T1 lines and get
shuffled out on other circuits - it's hard to get adequate mixing
except in rather large environments....  Recircuiting on the nets
will be left as an excercise to the reader.

I suspect the harder parts of the job may be doing the faces and 
voices right - anonymous voice conference bridges are ok if the
participants mostly don't know each other, but they're less useful
if people know each other and cops with computerized voiceprint equipment
may be eavesdropping (not common now, though computers and models of
the human voice are improving; I suppose voice disguisers may improve
from the kid's-toy quality to something better if there's a market,
or if computers with full-duplex soundcards become more common.)

Faces are harder, and they're not really a crypto problem -
how do you fake them well?  It's not too hard to do a "quayletool"
quality solution that generates moving lips in front of a static
picture, even timed with an audio feed, but that won't play too well
in the business world, and having the camera pointing at your calendar
or home page is only semi-useful.  If video-calling evolves on the nets,
there'll be a lot more need for speed-matching services, and it may
be that computer-enhanced video receiving for high-bandwidth users
will fund the technology development for face-simulation?
If so, maybe you can use it to start with fake stills instead of 
real ones?

				Bill