Ars Technica offers a tutorial:
First things first: the venerable set-top box, on its way out in the cable world, will make a resurgence in IPTV systems. The box will connect to the home DSL line and is responsible for reassembling the packets into a coherent video stream and then decoding the contents. Your computer could do the same job, but most people still don’t have an always-on PC sitting beside the TV, so the box will make a comeback. Where will the box pull its picture from? To answer that question, let’s start at the source.
Most video enters the system at the telco’s national headend, where network feeds are pulled from satellites and encoded if necessary (often in MPEG-2, though H.264 and Windows Media are also possibilities). The video stream is broken up into IP packets and dumped into the telco’s core network, which is a massive IP network that handles all sorts of other traffic (data, voice, etc.) in addition to the video. Here the advantages of owning the entire network from stem to stern (as the telcos do) really come into play, since quality of service (QoS) tools can prioritize the video traffic to prevent delay or fragmentation of the signal. Without control of the network, this would be dicey, since QoS requests are not often recognized between operators. With end-to-end control, the telcos can guarantee enough bandwidth for their signal at all times, which is key to providing the “just works” reliability consumers have come to expect from their television sets.