"... In PAGE 19: ... All the sessions are allowed to transmit packets for 2200 seconds, with the session starting times being staggered over the first second. Table1 shows the mean, maximum and the minimum value of the average transmission rates for three groups of sessions : the five multicast sessions, the five unicast sessions over L1 and the five unicast sessions over L2. The measurement interval is taken to be [200 sec, 2000 sec].... In PAGE 19: ... Each multicast session still uses the same LIF as in simulation 1, thereby tracking the more congested path of its two paths. We observe ( Table1 ) that on L1, all sessions (unicast or multicast) receive approximately an equal share ( 20 packets/sec) of the bottleneck bandwidth. There is less traffic on L2, hence more available bandwidth, however each multicast sessions is constrained to consume 20 packets/sec on all of its end-to-end paths.... ..."

"... In PAGE 20: ... Additionally, there are five unicast sessions over L1 only, five over L2 only and five over both L1 and L2. The results in Table3 indicate that bandwidth on the bottleneck link L2 is shared almost equally among the multicast sessions and the unicast sessions traversing both L1 and L2; hence each session receives about... ..."

"... In PAGE 39: ... Table1 shows the fraction of successful routes for 2-multicast sessions, routed through a 6-node, 8-link network, for the di#0Berent methods. In all cases, there was at least one solution, and the optimum algorithm always found it; the heuristics not always were able to #0Cnd the routes.... ..."

"... In PAGE 20: ... The value of has been chosen as 0:01. From the result of Table2 it is clear that max-min fair sharing of bandwidth both before and after the onset of additional congestion on L2. This is made possible because the LIF at each multicast source is able to always identify the most congested path.... ..."