Re: Simple Multicast - building a case for a BOF or WG

[Jon Crowcroft <J.Crowcroft@cs.ucl.ac.uk>]

 >>absolutely transparent to their OO world. 

i think this is often (not always) a case of putting the cart before
the horse - basically ,for each particular distributed application
that has multiple soruces, there are different requirements on
(and definitions of) reliability  ordering ,and synchronisatin of
delivery (and lots of other metrics)....the CORBA peeople would be out
of a job if we could define a single service that met all suvh
requirements......luckily we can't and also, they dont erally have
such a need - they have applications which have diverse needs and each
of which have specifal solutions

but this list is taegting network models anyhow, and not transport
ones-  though we want to have realistic requirements...
so what applicaiton requreiemtns have more than 1 or a small number of
sources right now...that are general and genuinely internet wide....?

 >>ISs requirement then: quasi-static shared multicast distribution tree with
 >>extremely low (--> zero) latency of multicast address allocation and
 >>extremely high dynamics of multicast groups (i.e. extremely short 
 >>multicast sessions). Note, that in fact almost every new group is a
 >>subgroup of a meta group (hence - quasi-static trees), and potentially
 >>any object of a meta group could become a sender (hence - shared trees).
 >>Note that CORBA-enabled hosts could be populated by thousands of active 
 >>objects (hence - dynamic address allocation is needed, not a block 
 >>reservation).

i am not sure i know hy yo uneed a group per object - not even the
LSVE (VR) people want that any more (after a close look at area of
interest managers and so on...)

 cheers

   jon