Ten years ago on Thursday (March 20), the very first cable-modem technology received the very first CableLabs seal of approval: It officially complied with the DOCSIS specification.
It really was a big deal. Despite its clunky name, DOCSIS was the industry’s first successful effort to create interoperable gear — which attracted more manufacturers, which lowered costs. A lot. What were $500 gadgets in 1998 are sub-$50 now.
(If you’re new to the scene, it stands for the Data Over Cable Service Interface Specification, and is pronounced “docksis.”)
DOCSIS, as a series of specifications, is entering its fourth chapter. Prior columns detailed what was in each version (1.0, 1.1, 2.0 and 3.0), and what’s new about the latest, 3.0 iteration of the specification. For that reason, we’ll assume you’re up to speed on channel bonding, wideband, IPv6, and the known world of DOCSIS 3.0.
Instead, this week’s translation will highlight two lesser known parts of DOCSIS 3.0, as a harbinger of what the spec might be and do in another few years. As in television.
IPTV (Internet-protocol television) remains one of those buzzwords best served with heaping portions of explanation. Because it still means different things to different people, it’s best to ask: What’s your definition of IPTV?
In this case, we’re defining IPTV as digital video, sent over the broadband IP (cable modem) path, to a screen. Not via a set-top box — not yet.
As of last July, the average streaming time was 2.9 minutes. That can only increase. And as it does, things need to happen to make sure video streams work well.
Right now, two small, geeky, and often overlooked parts of the DOCSIS 3.0 spec take the first steps toward an IPTV future for the technologies that began as cable modems: “source-specific multicast” and “multicast header suppression.”
Since “multicast” is the lingo denominator of both terms, we’ll start there. Short version: Multicast is the digital, Internet-y way to say broadcast. One to many. Its antonym is unicast: One to one.
Let’s say you and Bob live in the same neighborhood, and are wired into the same group of nodes. Unbeknownst to you, Bob tunes into Celebrity Sober House. A few minutes later, unbeknownst to Bob, you do too.
Let’s further say that you’re both watching the show over a broadband IP stream. Maybe you’re at your computer, or your cable modem is streaming the show over your Wi-Fi network to your laptop. Maybe you’re really hip and already have one of those broadband-connectable HDTVs.
Here’s what’s happening in the background, starting with multicast header suppression — which is sort of a data variation of “second verse, same as the first.” You’re getting a stream of packets. Each packet has a header, in the same way a paper letter has a written address. The idea with header suppression is to save bandwidth by sending the stream without the header packets, after their direction is established.
Here’s another way to look at it: Instead of saying “I want to send a packetized video stream to Harry. This is packet No. 1 to Harry. This is packet No. 2 to Harry. This is packet number No. 3 to Harry,” what if I said, “This stream is for Harry. Here’s packet one, packet two, packet three.” That’s header suppression.
Source-specific multicast is similar, but happens at the other end of the stream. This is “source-specific” as opposed to “any source.” So: You and Bob are both watching Celebrity Sober House. In “regular” multicast, the CMTS would set up Bob’s stream first (because he asked first). When you clicked to watch it, you would “join that stream.”
Fine. But north of the CMTS, out in the wild, wooly world of the public Internet, you and Bob probably hit the source separately — so there are now two separate streams coming into the cable headend. One for you, one for Bob.
A source-specific multicast sets up that stream once, from the source, to the CMTS. It’s an efficiency thing.
Right now, on its 10th birthday, DOCSIS is starting to tilt toward video — but the path is piled with debate. Will a teenage DOCSIS morph a redefinition of digital video distribution? The drumbeat grows.
But for now, feel free to take your data engineers to lunch this Thursday, or have a little party, or do something to show gratitude for what was and is a pretty big deal, as cable technologies go.
Stumped by gibberish? Visit Leslie Ellis at www.translation-please.com.
