Re: RFD:* (try #2)

Scott W Brim (
Thu, 13 Oct 1994 10:16:18 -0400

just fyi, I sent this to rem-conf recently...

CU-SeeMe reflectors do use unicast at the moment, and if they are set
up wrong they can send multiple streams over a particular circuit.
People set up "nets" of reflectors all the time to minimize this.
Since each "reflector net" has its own infrastructure, it isn't as
complicated as setting up the mbone.

While I'm at it, I should explain a few concepts we have for future
implementation. Reflectors will multicast to each other in "reflector
nets" authorized by their administrators. Each "reflector net" will
use its own range of multicast groups, for control and data flow in
each "conference". Thus traffic per conference will be minimized (to
the extent the mbone traffic is minimized ... hmmm :-)), but we'll
still have the administrative controls those network managers love, as
well as an easy place to work on things like personalized information
flows. We're going to hang on to the reflector concept for a while,
even when reflectors and nv are seamlessly interoperable and when
everything in the world supports multicast. It isn't in keeping with
the "distribute everything ultimately" model which we all find
extremely aesthetic, but it offers tremendous opportunities for
exploration. Of course none of this will interfere with users
connecting directly to each other.

As for saving the world from meltdown ... well, I'm hoping that
providers will charge what they have to to provide acceptable service
-- that if they are being used heavily enough that they can't deliver
what they have promised, then they will also be bringing in enough
money to upgrade their infrastructure so that they *can* deliver. I
don't have any answers for how to upgrade every school in the world so
they can have good conferencing connectivity -- we're still trying to
put as much information as possible in as little bandwidth as possible.