Visit Your Local PBS Station PBS Home PBS Home Programs A-Z TV Schedules Watch Video Donate Shop PBS Search PBS
I, Cringely - The Survival of the Nerdiest with Robert X. Cringely
Search I,Cringely:

The Pulpit
The Pulpit

<< [ Bursted Not Busted ]   |  Hide and Seek  |   [ There's No Show Like an Old Show ] >>

Weekly Column

Hide and Seek: VoIP Users Strike Back Against Network Providers, but Will It Be enough? No.

Status: [CLOSED]
By Robert X. Cringely
bob@cringely.com

One characteristic of the Internet that has remained a constant for as long as I have been online is that there is never enough of it. If you have a 300 baud modem, you long for something faster. If you have a 9.5 megabits-per-second broadband connection like I do (two cable modems and a DSL line ganged together), you still want something more. Maybe my connection is fast enough (barely and only just until something faster comes along), but now, the rest of the network shows its limitations through overloaded servers and bad routes. The answer, we're told, is fiber-to-the-home, plugging fast-Ethernet into the back of your computer, TV, toaster-oven, etc. But that isn't the answer at all, because a bigger pipe is never big enough, especially when it has to function in a complex communication culture that quite deliberately does some things a lot better than others. What is actually needed is a better understanding of how to really optimize the network so that everyone is better served -- something that is coming not because we're being altruistic, but because the alternative is digital gridlock. And that's why most streaming audio and video are quite naturally reverting to an older paradigm -- downloading and storing.

"The heart wants what it wants," said Woody Allen, and it is true: we all want better Internet service all the time. I impatiently watch my e-mail download. My E-MAIL! Why should I care if it takes 15 seconds or 17 seconds? Yet I do. We all do. It's what keeps us buying faster PCs and new Internet connections. So one given has to be that nobody is going to voluntarily reduce his bandwidth usage. Even those of us who run proxy servers only inadvertently save bandwidth. Our real goal is to make web pages load faster.

Streaming is wonderfully selfish activity. What we are asking the network to do is to open a high-speed lane between our screen and some source of video content, then drive that data to us as fast as possible. The fact that there can be as much comprehension created by a 100 kbps stream as a 300 kbps stream is meaningless. We just want a bigger picture with HDTV-quality, the new Holy Grail. Alas, streaming doesn't work very well for the most part because our networks are deliberately overloaded, or to put it more properly, they are over-committed.

I have a couple Comcast cable modems that were very recently increased in speed from three megabits to four megabits-per-second. I didn't ask for that increase, Comcast just gave it to me. Whoopee! But what does it mean? Does it mean that Comcast increased its Internet bandwidth capacity by a third? No. It means they increased by a third the amount they oversell their existing bandwidth capacity, running it up to around 200-to-1. In other words, for every 200 paying customers, Comcast provisions about four megabits-per-second of backbone capacity. If, some Saturday night, we all decide to get online and start watching music videos, well it simply isn't going to work, but that's a risk Comcast in this case is willing to take so they can say they are giving us more for our money.

This phenomenon plays back to my recent columns about how telcos and cable companies are reacting to VoIP competition. Increasing maximum bandwidth while doing nothing to increase backbone capacity just pounds another nail in the casket of "best effort" service.

Fortunately, against this dismal landscape we have a couple positive events. For one, people like to own things, or at least to think that they own them, which is the inherent strength in iTunes versus Napster. If people like to own stuff then they intend to keep it, which means storing it and indexing it in some way. And finally, nobody has time to do things like entertainment in real time unless they plan in advance, as in, "Let's go see a movie tonight, Honey." That exchange may feel spontaneous, but compared to video-on-demand streaming, it is glacial.

The truth is that almost nothing on TV today is truly spontaneous, with the possible exception of the odd natural disaster or slow speed chase. Even when the President gives a speech, the text has been in the hands of reporters for hours. The news stories we see on TV are generally hours or days old by the time they reach us. Those of us who get much of our news online are even dismayed at the delays in that medium. Google News, for example, is by definition reactive so stories don't reach the top of the queue until hours after they are broken.

Then there is that moment in the Big Game or the Big Show that you absolutely can't miss except your significant other wants to talk about the kids or the garden or the armed fugitive climbing over your back fence -- events totally inconsequential in comparison to the Big Catch,and yet, you know somewhere in your dim brain that your relationship depends on paying attention not to the TV, but to the person across from you just at that key moment. That is, of course, TiVO Time.

None of this is going to go away. Networks will get faster, but bandwidth requirements for content will grow as well, which means download-and-play is the future of streaming.

But don't mistake me: Download-and-play can happen in less than real time if the bandwidth is available. Download-and-play can begin showing the content before it has all been downloaded. Many users won't even think there is a difference between streaming and downloading, except that at the end of a download, you have the whole show or song on your hard drive.

Bit Torrent and Torrent-like services fairly demand download-and-play to work, and I can tell you right now that the cost-effective replacement of broadcasting by Internet distribution absolutely requires something like Bit Torrent, once the content producers are a little more comfortable with the idea.

Embracing downloading, rather than streaming, allows much more efficient bandwidth utilization at little or no perceived cost to the user. It also can allow closer attention to services that actually require low latency, like voice and video chat and VoIP. Toward that end, a game of hide-and-seek is presently unfolding in which mainly Open Source software developers are trying to find ways to make their VoIP and chat packets look like anything but.

There are several different methods being proposed to elevate packet priority in QOS networks. Some take advantage of the fact that QOS classification is often handled at beginning of session and not with each succeeding packet. Some embed hitchhiker content into existing traffic streams. Special routing programs unblend the unauthorized traffic and forward it once it reaches a friendly gateway within the larger provider's network. That's a lot of work, but then there is a lot of money at stake. You'll find references in this week's links.

Despite all this effort on both sides, I have to think that low-latency traffic is ultimately destined to travel in IP tunnels, and IP tunnels (VPNs) steal bandwidth from everyone else. Not good.

This all comes down to ownership of the last mile, just as it has ever since the breakup of the Bell System in 1983. And it isn't just in the U.S. The same shenanigans are taking place in Europe. But we are rapidly approaching a time when there may be a viable last mile alternative specifically intended for such low-latency applications. Two years from now, the balance may have shifted completely, both in the U.S. and abroad. But I'll have plenty of time to write about that next week.

Comments from the Tribe

Status: [CLOSED] read all comments (0)