I baldly claim here that IP traffic protocols do not suit voice traffic and other streaming sound and video. No one who is listening to music, to enjoy it, will tolerate occasional dropouts. It is not the cost of delivering the bits, it is the inability to provide more than 99.9% probability of getting the packets thru in the presence of competing fractal traffic.

Cisco is pushing IP telephony and they have some protocols relating to their Tag switching which can be extended to make IP support voice and other streaming.

Head bone — Knee bone

Recently SONET framing was the dominant bottom level format for data over optical fiber. This is because the telcos had bought enough SONET hardware to make the framing part cheap. The other part of SONET is time slot interchange which is used to switch constant bandwidth 64Kb voice streams. ATM uses the former part but not the latter part.

There are a variety of QoS options for ATM. Some of them were developed to support voice but take advantage of compression so as to use the resulting excess bandwidth, at least if competing traffic is POTS like.

DSL seems to be the last mile for ATM. I think that ATM is in a position to deliver what IP promises for AV streams. I imagine making arrangements to listen to some music for a while thru the web (over IP) and then switching to an ATM SVC to actually carry the stream.

It is not clear to me that ATM provides appropriate adjudication between a stream and bursty traffic when together they compete for the entire capacity of the DSL part of the SVC.