The current democracy of internet packets will disappear in favor of unfair, Quality of Service castes like high-priority, urgent, express, and other superlatives. The addition of QoS ("kwoss") to standard, random-length* IP packets will begin to enable the multimedia dataflow that our anarchic internet demands. This will hopefully achieve the same results intended by cell-based distribution schemes like ATM, which enforces very small packets to lower both the expected packet delay and the packet-delay variance. If the worst-case delay variation is kept to a minimum, users get the most immediate information and require the least amount of local buffering. The small packets (cells) are harder to route and consume large amounts of overhead bandwidth, so the hope is that, with a sound baseline in routability and managability, IP+QoS can serve as a single vehicle for all multimedia payloads. It remains to be seen if the IP+QoS approach will be acceptable directly or may lead to the necessity of segmentation and re-assembly of packets (or moves toward encouragement of smaller packet sizes) in order to meet guaranteed low-latency values.

Another possibility, aside from homologization to IP+QoS, is that via DWDM and other high-end optical communications possibilities, we simply 'open a new channel' for voice and data services. We would then have an overlay of public voice and public video networks on the current data internet, and each packet gets switched based on type to the appropriate infrastructure. The multimedia traffic would then be aggregated where possible and segregated where necessary to meet the bandwidth and latency goals. This would be accomplished at a cost of channel allocation and management overhead as well as requiring frequency-nimble switches (namely, a switch that can connect any input stream to a particular DWDM channel or color.)