that you overestimated the demand for live IP video, or do you think that your timing of the demand was just off? I'd say the millions who watched March Madness and subscribe to Major League Baseball's web video might disagree with your assessment, and certainly delivering multicast video is very much a part of the strategies of the telcos and cablecos insistent on continuing down the path of a non-innovative IPTV, part of the larger shift to the ingrained service provider model that (1) they hope to ensure their survival and (2) quite probably, is the underlying motive behind the tiered Internet initiatives. With that, questions, which you may or may not be able to answer as a former network architect with focus on bandwidth-hungry applications: could certain re-engineering in the core overcome the limitations of unicast peering while maintaining dumb long haul pipes, why isn't throwing more bandwidth at the problem a practical solution, and can't telcos with VoIP offerings already guarantee QoS to several nines without packet prioritization?

"Round and round, round we go." - Tupac

The crux of the problem in 2000 was that nobody had a reasonable business model. We had salesmen calling on NFL franchises with the following proposition: "You, Mr. NFL Owner, pay me $1,000,000, and I put your FOOTBALL GAME on the INTERNET!" There were a few event promoters who were willing to cut that kind of a deal in the year 2000 (both political parties, a few entertainers, a couple of boxers and soccer stars), but not many.

A single, simple principle rules Hollywood and the television universe: Content is king, technology is its footservant. Telecoms and software companies are slowly figuring this out, and a surprising number still don't get it. Microsoft has spent millions and millions of dollars for something like a decade trying to get a foothold in the digital entertainment business, and they still have basically bupkis to show for it. They'd love to have the success Apple has had (through leveraging their history with the creative community as well as the Pixar relationship with Disney), and in turn, either of the two major telecoms I've worked for in the last five years would love to have what limited success Microsoft has had.

I think multicast is likely at least part of the reason why you're seeing new service offerings these days concentrated in the wireless space. Every video packet that goes over the Sprint PCS network is going to a Sprint subscriber; every packet that goes over the Verizon network is going to a Verizon subscriber; and similarly for Cingular and T-Mo. None of that content has to cross peering points, which means that it can all be multicast or cached within the network near the subscriber, as appropriate.

Telecom, in general, is in a bad way. The financial foundation of the industry, local wireline telecom, is such a bad value proposition for money spent these days that it's little wonder that customers are abandoning it in droves. The local telephone operating companies have for the most part dried up and blown away; a couple of them have succeeded in diversifying into other product offerings, but in almost every segment of the industry, internet-induced innovation drives revenue potential through the floor, and promises to continue to do so for years to come.

While POTS subscribers continue to flee Verizon, SWC/AT&T, Qwest, and BellSouth in droves for the likes of Vonage and Cox, at the same time Verizon, Cingular, Qwest, and Sprint face the prospect of losing billable network minutes to a new generation of handsets with integrated WiFi enabling you to sit at your local free-hotspot coffeehouse chatting away over the internet without paying your cellular provider a cent. Steve Ballmer is already making ominous vague threats about the killer telephone app Microsoft has tucked away in Redmond awaiting release.

Perhaps worst of all, even WiFi itself is facing obsolescence, soon to be superseded by new wireless wide area networks bound to make it more difficult than ever for T-Mobile and Boingo to steal business from cellular subscribers hanging out at Starbucks, McDonald's, Borders, and Barnes & Noble.

Meanwhile, powerful people in Washington are making threatening noises about how the big bad telecoms are taking unfair advantage of their monopoly power, insinuating the possibility of government regulation of telecom trunks. All these companies are in business to make money, and their traditional revenue streams are drying up. I think the future of the internet is at least as threatened by the specter of bankruptcy of backbone carriers unable to meet their operating costs as it is by the prospect of cost recovery for innovative engineering.

To address your questions: the "limitations of unicast peering" amount to a multiplier of several thousand fold in aggregate bandwidth consumption entailed in transmitting a live video stream to a large audience. Placing tongue firmly in cheek, that's pretty hard to handwave away.

As I pointed out above, IF you can control every packet you push over your IP pipe, and IF you use a low enough fraction of the available bandwidth, and IF you have the best networking operations people you can find, and IF they do their jobs perfectly, every day, you can achieve some pretty amazing things over a DWDM fiber. That's a lot of ifs, and ultimately the problem is that the definition of public internet is a pipe on which you DON'T have control of all the packets, and you CAN'T dictate that only, say, one percent of the available bandwidth can be used, and once in a while, SOMEBODY's packets are going to have to wait, buffered up in a router somewhere, while somebody else's packets use the network. If you want, you can build optical networks on which nobody ever has to wait, on which everyone has all the bandwidth they need; but that's not an IP network. By definition, an IP network is what we call "store and forward"; each packet travels from source to destination via a series of hops, between which it's buffered at an intermediate node somewhere. Under conditions of negligible bandwidth contention, it will behave in many ways like a network with dedicated bandwidth, but it's not one.

Which brings me to VoIP. Every audio or video stream that traverses the internet does so as a series of packets. These packets, like all the other packets, hop across the network, stopping momentarily at the intermediate nodes we call "routers".

In order for an audio signal to be digitally sampled, encoded, and packetized, it has to be buffered for some amount of time in memory at the sending end. Each of the packets makes its several hops across the network, and at the receiving end, they have to be buffered again and reconstituted into an audio signal. This takes time; if you've watched streaming video, you're used to waiting ten, fifteen, or twenty seconds for a stream to buffer up on your desktop.

If any of these audio packets fail to arrive or arrive out of order, you will hear a glitch in the decoded audio. You can ensure that the packets arrive in order, by tunneling your audio stream over the transport protocol we call TCP; but that doesn't prevent packets from arriving out of order, it simply means that packets that arrive too soon sit around and wait for late packets to arrive before they're presented to the listener.

If you use TCP, the cumulative delay you experience in waiting for late packets adds up, and if you're trying to carry on a two-way conversation, it either slows down to an annoyingly slow pace, or it breaks up because you can't keep track of what the other guy is saying. So we use another protocol called UDP, which does the best it can to deliver all the packets in order, but if something goes wrong, rather than waiting for late packets to show up, it just drops them on the floor. You hear a glitch in the audio.

This happens all the time, on anyone's VoIP circuits. Nobody gets "several nines", but fortunately, nobody NEEDS "several nines". You've been listening to 4 kHz POTS audio for so long that you don't notice glitchy audio from dropped packets; in fact, the thing that bugs me worst about my VoIP line is that when nobody is talking (and thus no data is being sent over the network), it's TOO quiet. It sounds like the person on the other end hung up, or something.

If you're sending hidef video, or something that *needs* QoS, the way you achieve that on an IP network is to assign those packets a high priority. Other packets, like people's instant messages, Napster downloads, and VoIP conversations, sit around in the router and wait while your packets get moved. That's the magic of IP; that's how everybody's data gets moved without anyone having to pay for the peak bandwidth they use. If you absolutely, positively have to have your data at the other end of the network in a fixed amount of time, on the other hand, IP's not the solution for you; you want an ATM circuit.

Unfortunately, those are a good deal more expensive.

Methinks we're getting semi-off topic here, but that's good. Net neutrality is a convenient, though important (IMO) part of the larger issues you write about.

1) The wireless world is still a hugely different beast than wireline. You can do a lot less with 3G or present EVDO cards than you can with FTTP/FTTH. Is it also not far-fetched to add to that the nature of the beast, that they also largely control the handset traffic comes in on? It's complete end-to-end control, every carrier's dream. Very rarely can I just plug and play like I can with a TV, computer or landline phone.

2) Cost recovery: I'll go ahead and declare small local TDM wireline largely dead. RBOCs who have feet in quad-play rule the roost for now, anyway, and everybody else (Clearwire, SprintNextel) is just the newfangled version of the leftover LEC crowd, at least until the juggernaut that may be IMS comes full circle. Surely the big guys get huge discounts from Cisco/Juniper/Alcatel/Nortel etc. on network equipment. They also get tons of money in recovery for degraded equipment (I've heard Verizon got significantly more back last year than the total spent on its costly fiber build-out initiative). Yes, these guys are dinosaurs and they don't really yet understand the monetary consequenses of real competition, but they're still right now heavily subsidized and have tons of subscribers to rake in billions annually. The coming platform war that you speak of, though... you know there's little incentive to bring WiFi phones to our shores, people still want cable-like TV (though there will be naysayers), every tech thus far has found its own niche (WiFi, wireless mesh, EVDO/OFDM, fiber, etc.) and why isn't just paying for Internet access enough? They've always sold just access and some added services, now they'll just do it on a different model. These guys won't die, they'll change with the times, as they are already doing by moving to IP.

3) You mention the waiting time in routers, but we now have routers by Cisco, for example, that can handle an array of OC-768s adding up to almost 100 Tbps. That's a ^&*#load of traffic and would certainly minimize any latency caused by the router itself.

4) [Disclaimer: here's the one where though I think I know what I'm talking about, in reality I'm sure I may have something technical mixed up, especially as regards specific protocols and when they are used... not a carrier network buff] First, the IETF already has infrastructure recommendations so that we can all get along. Additionally, we can add to that: what about network optimization: a slew of monitoring and management appliances/software, deep packet inspection, throwing more metadata on packets, the continued development of additional protocols like IGMP, new techniques to do multicast packet forwarding using MPLS, and protocol checking/optimization in general. I sure as hell can't afford Akamai for personal use, but making telcos invest in at least a portion of this stuff would make it a little unnecessary anyway. Wouldn't all those lessen the blows and give carriers every opportunity to make multicast video specifically work as well as possible without real deprioritization to other types of packets?

5) With 100 Mbps to my home and the above two technical elements in place, I'd think any buffering or dropped packet issues would be terrifically minimized toward a point equivalent to the "blank screen" time it takes when I change channels on cable TV. We put up with way more than that every time we now load a web page, say, even The Agonist.

"Round and round, round we go." - Tupac

ok, let's talk about net neutrality.

My contention is that if the words "net neutrality" mean anything, they mean "state control of telecom trunks". This would be one thing in a society with a tradition of regarding telecom infrastructure as a public investment, but our society is quite the opposite; we are so firmly ideologically wedded to private enterprise that the most we could be persuaded to do was provide mild incentives and legal context for private companies willing to undertake the investment. (Even those societies - mostly in Europe - that have traditionally had state-owned telecom have almost universally privatized those enterprises by now. Some of them - such as Deutsche Telekom, which has become one of the big cellular operators in the US, and the largest provider of pay-wifi - are quite viable.)

I'm not a lawyer, and I don't really want to get into a legal discussion; my point in raising the issue is simply to point out that there are currently no federal regulations dictating any kind of legal policy regarding internet routing or peering. Legally speaking, the engineering standards of the IETF are suggestions; that might actually be a little too assertive a characterization, since the acronym RFC literally stands for "Request For Comments".

The internet is a public resource only in the sense, and to the extent, that the owners of the private property that constitutes it have donated it for public use. My broadband connection is not a public resource; it is a service I buy from the cable company. If I choose to leave my wifi access point open, it's not because the public has a right to sit in my back yard and use my internet connection; it's because I'm a nice guy, or because I'm too lazy to deal with encryption.

"Ah," you say, "but the cable plant is a public resource." In some places, that's true. I happen to live in one of the rare places where the TV cable plant is actually meaningfully managed as a public resource. There are two cable companies serving my community, and I am free to subscribe to either of them. Most places, though, don't give you this option; most municipalities have allowed themselves to be railroaded into one cable TV monopoly or another. Since local cable franchises are bought and sold as private property, and since most municipalities have essentially zero likelihood of attempting to reassert control of their local cable franchises, they might as well be public property (and the telecoms are lobbying Congress for essentially that, so that they can cherry-pick local TV markets without having to worry about the remote yet troublesome threat of assertive local control).

At any rate, there's the assertion that the public owns the right-of-way over which telecom trunks are laid. Again, I'm not a lawyer, but I can tell you that at least one of the big telecoms that I've worked for grew out of the side of a railroad. The railroad was being paid by a telecom to lay optical fiber in its railbed; it decided that for every cable it laid for a telecom, it would lay one for itself. Once the owners of the railroad had enough dark fiber in railbed to constitute a credible national network, they started a telecom company called Qwest.

Although the official version of their history is quite different, I've been a customer of Sprint since very near the beginning; they also started out as a wholly owned subsidiary of a railroad. (Later on they were partly owned by GTE, and later yet they acquired the minor independent United Telephone, which is where the official corporate history now traces its origins, but at the beginning the "SPR" in "Sprint" stood for "Southern Pacific Railroad".

So you can claim that telecom trunks are a public resource, but the reality is that they were created and installed by private investment. That won't necessarily stop Washington from taking them over, but in a fair society it ought to.

Since local cable franchises are bought and sold as private property, and since most municipalities have essentially zero likelihood of attempting to reassert control of their local cable franchises, they might as well be public property (and the telecoms are lobbying Congress for essentially that, so that they can cherry-pick local TV markets without having to worry about the remote yet troublesome threat of assertive local control).

Whoops. what I meant to say was, "they might as well be PRIVATE property..."

I see that I've largely failed to address your points, so let me take on one of them directly: the IETF does in fact have engineering recommendations "so that we can all get along".

All the bluster about "net neutrality", though, is carrying the message to the telecoms: don't innovate. Don't implement those IETF recommendations; don't set up multicast peering; don't do anything different from the way you've always done it. Don't make it possible for everyone to get along.

It's largely as if you were to go to Intel or AMD and say, "We don't want your new generation of multicore CPU architectures. We don't want you to build multicore CPUs, because the clock speed isn't as high as it was on the Pentium 4; we want you to keep raising the clock speed, power consumption and heat dissipation be damned. We feel so strongly about this that we want Congress to forbid you to make such chips."

Last time I sat down and thought about it, optical networking technology was moving at a pace even faster than Moore's Law; DWDM transceiver technology was moving so fast that manufacturers were making new-generation equipment not twice, but four times as fast as the previous generation. Every three years, the potential bandwidth of a strand of already-installed dark fiber was growing by a factor of four, just by lying where it was in the ground.

I have no idea how long the technology can move at that pace, or indeed whether it still is. I suspect that the demand for optical network transceivers may not be what it was in the late '90s, when all the backbone carriers were overbuilding facilities at a ridiculous pace. But the important point is that the available bandwidth is not growing, and cannot grow, as fast as the demand for bandwidth potentially can; and it will not grow unless the carriers can show a return on investment.

The rest of your arguments, by the way, are mere handwaving. You're essentially saying that, although UDP/IP is an unreliable, packet-losing protocol that doesn't provide high-quality transmission for bandwidth-intensive applications, we'll make it up in volume.

Why is it mere handwaving to suggest that bandwidth and protocol optimization is an answer to problems? It's certainly an answer for lots of big businesses and for countries like France and South Korea.

"Round and round, round we go." - Tupac

It's not a message not to innovate. You've taken that chestnut hook, line and sinker. Their initiative to innovate is the billions we pay for Internet access every year and the money I pay in taxes for the USF and the tax money they get for degraded equipment. Hell, if there wasn't an impressive uptick in broadband speeds, if they'd owned up to promises in the 90s to build fiber into communities back then or at least stayed on their timelines (for example, fiber throughout New Jersey by a few years ago), and if they weren't trying to compete with the Vonages and Googles of the world, then maybe I'd agree.

It's nothing like going to chipmakers and saying that. Instead, going with the computing paradigm, it's like AMD or Intel optimizing their chips specifically for one application, but not for others. I don't want that, either, but with the telecoms its worse because chips have never been thought of in that way and the telecoms have long been known to try and block or stop services (including trying to block Prodigy, the Madison River case, and on back for years). Hell, that's the reason for the original common carrier laws in the first place way back when.

As for wave division multiplexing, that proves my point more than it does yours. Optical networking technology has moved at that rate not because of a lack of net neutrality, but because telecoms knew people want and increasingly need lots of bandwidth. That multiplying potential can solve any problems that you say are constrained by net neutrality. You yourself say available bandwidth is growing just by sitting in the ground, and add to that that most of the strands of fiber aren't even being used yet. Of course it can't grow as fast as demand "potentially" can... I'd love to have 15 gbps throughput today, ut I know it's not feasible. As for a return on investment, just look at BellSouth's latest high-speed DSL numbers for an answer.

"Round and round, round we go." - Tupac

I don't have access to anybody's router statistics, but I can tell you from firsthand experience that video eats up bandwidth at a rate that's difficult for a layman to comprehend, particularly when it's passed over an IP network in the form of unicast streams.

If you have typical cable TV broadband in the form of a DOCSIS cable modem, all your downstream data, as well as all your neighbors' downstream data, is multiplexed over a single cable TV channel. The typical cable system contains something like 120 of these channels because it needs the bandwidth to pass all that video.

Since the cable TV network is dedicated to broadcasting video, it only has to send one copy of each channel. Imagine if, instead, it was necessary for the cable company to dedicate an entire cable to every subscriber; instead of having a single coaxial cable strung along telephone poles on your street, there would be a bundle of cables the size of your arm, or possibly the size of an HVAC duct.

This is, in effect, what you have to do when you're passing video over a public IP network. Every subscriber has to receive a separate copy of every program. Imagine tens of thousands of coaxial cables running out of ESPN headquarters in Connecticut to living rooms across the continent. It's no wonder the telecoms are getting tired of dedicating all that bandwidth to video, when with a little technology they could divide that burden by three or four orders of magnitude.

VoIP, by comparison, is a tiny problem. But try to look at it from the point of view of the telecoms. Over the decades, their business has been that of providing measly 4 kHz audio connections over twisted pairs of copper. They've had a government-awarded monopoly, true, but in return they've had to operate in a fairly restrictive legal environment.

All of a sudden, somebody with a switch connected to the TDM POTS network and an internet trunk sets up a gateway between the two. People who already have internet access at home don't see any reason why they should pay the artificially inflated price of regulated telecom when they can just use their internet pipes, and they cancel their POTS service. Naturally the regulated telecoms are going to see this as unfair competition, and they're going to object to passing VoIP traffic over their trunks for free.

They're naturally going to see it as unfair that Vonage, Skype, and Packet8 are allowed to provide telecom service at bargain rates while not being required to provide emergency 911 service. (They're not, any more, by the way; Congress took care of that last year.)

More in a few minutes.

I am suggesting that the telecoms are not interested in a single source datastream because they see a new revenue model but it is only indirectly related to bandwidth concerns. As you say, VOIP threatens to rob the telecoms of much of their revenue while video eats up their bandwidth. Solving the bandwidth problem does not save them from being eaten alive by VOIP.

How will the telecoms make money with the their proposed controls? It would be very hard, though not impossible, to set up an infrastructure for tracking and charging using a single source stream with multiple destinations using peering. I know because one of them would probably want to hire me to figure that out.

Protocols like SIP are a negligable part of VOIP bandwidth which, as you say, is small compared to what video uses. For example, SIP simply sets up a VOIP connection and potentially video connections but does not transport the data. However, if you monitor or control the SIP protocol and any protocol like it then you will have control over VOIP and eventually over video since SIP or something like it may eventually be used to set up video connections too. You could charge per call, video session, or per packet for either. A law that is very vague could allow the telecoms to gain this kind of control which IMHO would endanger the Internet's future much more than charging simply for bandwidth.

We shall not try to make these people or any of their children into philosophers or men of learning or men of science. - General Education Board Letter #1, 1906, Rockefeller Foundation.