On Mon, Oct 24, 2011 at 10:42 AM, John Leslie <john(_at_)jlc(_dot_)net> wrote:
Mikael Abrahamsson <swmike(_at_)swm(_dot_)pp(_dot_)se> wrote:
On Mon, 24 Oct 2011, John Leslie wrote:
150 milliseconds is a real challenge to accomplish worldwide, though
it's quite achievable within one continent. I expect IETF folks could
learn to work with 250 milliseconds.
Are these numbers RTT or one-way?
I'm embarrassed to admit I don't know which Brian Rosen meant: perhaps
he'll elaborate. He may well have meant one-way delay plus codec delay
plus application delay.
The speed of light in optical fibers (index of refraction ~ 1.5) means that
the one way time from any place on Earth to its antipode must be >~ 100
milliseconds, so I think that the above must be one way times.
300 to 400 milliseconds to the antipodes and back (i.e., RTT) is pretty
realistic (say, US to Australia*). To that has to be added codec delays
(each frame of 30 fps video represents ~30 msec);100+ msec one way video
codec delays are common. If you add all of that up, you get enough latency
that it begins to be noticeable, even in a formal meeting, for links such as
US-India and US-Australia.
*The closest to truly antipodal pair of places I know of in common use are
Hawaii to South Africa. If anyone has measured RTTs for that I would be
curious to know them.
According to figures I've seen in other contexts, most people are fine
with 400ms RTT (this is a quite common delay just talking mobile
phone-to-phone even in the same city),
I'm pretty sure what I've observed mobile-to-mobile exceeds Brian's
but people really start to notice around 500-700ms RTT. 1 second RTT
is really noticable, but still workable with some practice.
_I_ certainly notice before 500 msec RTT-plus-codec.
I don't think I agree that 1 second RTT-plus-codec is workable in
groups where any of a half-dozen folks might speak at any time.
It's hard to have a heated argument over more than 400-500 ms RTT
so it depends on what kind of discussions are to be had :P
It wouldn't be IETF without an occasional heated-discussion!
Ground/sea based fiber optical cable networks rarely give more than
500ms RTT, so anyone fairly well connected to the worldwide Internet
via ground based infrastructure should be able to participate with
less than 1s RTT including encoding delays etc,
There's no reason why ground/sea based fibre needs to exceed about
200 msec RTT; but buffer-bloat does cause this sometimes. In practice,
business-level Internet is likely to add 100 msec to this, and cable
Internet can add considerably more. :^(
at least if the system is located at the same place or fairly close
to the venue
(I'm guessing you mean a single-central-server through which all
(at least so the signal doesn't have to be bounced half way around
the world before it's sent to the final > destination).
I don't honestly know how flexible the various vendor systems are
in that respect. I would like to believe they are capable of more
intelligent switching than that...
John Leslie <john(_at_)jlc(_dot_)net>
Ietf mailing list
Ietf mailing list