🚀 go-pugleaf

At 9:21 AM -0700 10/8/98, Andy McConnell wrote:
>Hearing this discussion, it reminds me of another reason NOT to charge for
>it - reduced backbone bandwidth.  Why discourage your customers from using
>a higher-quality service that costs you LESS b/w to provide ?

The reduced bandwidth argument is a little bit tricky.  I.e., "reduced" in
comparison to what?  If you compare to the load that n unicast receivers
would place on the net, there is an obvious reduction.  But, if you compare
to the load that 1 unicast source can place on the net, there is an obvious
increase.

This suggests that receiving multicast might be free, but sending it might
cost extra.  On the other hand, the additional sending cost probably gets
passed on to the receivers in some fashion anyway, e.g., by making them
view advertising.  If the ISP doesn't collect it, the sender probably does.
Not sure how a transit carrier fits in economically, since it is both
sender and receiver (from the point-of-view of different parties it
contracts with).  Perhaps it all balances out.

Also, important not to neglect the additional routing state that a
multicast session creates.  Since multicast addresses cannot be aggregated,
I guess the effect of an active multicast group is roughly the same as a
new network attaching to the Internet for a while?

Tim

Tim Dorcey
iXL-Los Angeles
tdorcey@ixl.com
(310) 235-3928