Internet Myths

Among my missions in this life is the chore of explaining networking in general and the Internet in particular to policy makers and other citizens who don’t build network technology for a living. This is enjoyable because it combines so many of the things that make me feel good: gadgetry, technology, public policy, writing, talking, … Continue reading “Internet Myths”

Among my missions in this life is the chore of explaining networking in general and the Internet in particular to policy makers and other citizens who don’t build network technology for a living. This is enjoyable because it combines so many of the things that make me feel good: gadgetry, technology, public policy, writing, talking, and education. It’s not easy, of course, because there are a lot of things to know and many ways to frame the issues. But it’s possible to simplify the subject matter in a way that doesn’t do too much violence to the truth.

As I see it, the Internet is different from the other networks that we’re accustomed to in a couple of important ways: for one, it allows a machine to connect simultaneously to a number of other machines. This is useful for web surfing, because it makes it possible to build a web page that draws information from other sources. So a blog can reference pictures, video streams, and even text from around the Internet and put it in one place where it can be updated in more-or-less real time. It enables aggregation, in other words. Another thing that’s unique about the Internet is that the underlying transport system can deliver information at very high speed for short periods of time. The connection between a machine and the Internet’s infrastructure is idle most of the time, but when it’s active it can get its information transferred very, very quickly. This is a big contrast to the telephone network, where information is constrained by call setup delays and a very narrow pipe.

Both of these features are side-effects of packet switching, the information transfer system used by the Internet and all the stuff that rides on it. Packet switching isn’t hard for people of reasonable intelligence to understand by reference to postcards and the like. Engineers can generally take such an audience on a productive trip through the essentials in a few minutes, even relatively unskilled ones. It’s not that hard.

Some of the people who make a living explaining the Internet to the mass audience get it all wrong, however. Most of these people are lawyers or law professors, which is kind of mystifying because they’re bright people and really should know better. One such example I encountered this week is an article in The New Republic by Tim Wu comparing the Internet to movies, radio, and TV. I’ve met Tim and find him to be a bright person, certainly brighter than many of this para-technical adversaries, so it’s odd that he would make statements that are simply ridiculous on their face. Wu invents a fictitious element:

History does not always repeat itself, to be sure; and it is possible that the Internet is different. Its most fundamental feature, after all, is a radical separation of distribution and content, which is the very opposite of the NBC/Paramount system…

The Internet’s wall between content and distribution also creates considerable power for those who can harness it. It is little appreciated how dependent the Internet’s business models are on a neutrality in the infrastructure.

How radical it would be, according to Wu, for someone to build a radio that could carry spoken words without regard for their language, message, political orientation, or emotional expression. Who could possibly imagine a movie theater capable of showing content produced by Warner, Sony, or Four Wall Productions? How revolutionary it would be if we had a transportation system capable of handling trucks laden with either the Wall St. Journal or the New York Times, instead of the tightly interconnected system we have today!

Where does this nonsense come from, and who can possibly believe it?

Wu is probably mislead by the language that engineers use when describing network structure (or “architecture” as we like to call it.) At the dawn of packet switching, we mistakenly guessed that each problem in the delivery of a packet could be solved once and for all at one point in the system and if we could identify these points we’d be on the right track to developing a general theory of network design and optimization. The embodiment of this idea was the Open Systems Interconnection reference model, which decomposed networks into seven functional layers from application (at the top) to physical signaling (at the bottom.) Somewhere in the middle a line was drawn between content and transport, or more pertinently between representations of content and systems of transport. But this structure was simply an abstraction that can be applied to any system of information or transport just as easily as to Internets.

Movies represent information as sounds and pictures on a strip of film, and films move from studio to theater in trucks. Replace the film with a disk file and substitute fiber optic cables for trucks and nothing changes in the realm of structure or description.

The Internet moves digital information, and it does so by copying bits and then discarding them. Bits move quickly and inexpensively because the ratio of information to mass in a digital system is very, very high. Walls of separation are fine and good in the realm of law, but they have very little to do with technology unless we want them to.

The Internet differs from old media technologies in terms of speed, cost, and reach, but not in structure. Now you know why.

12 thoughts on “Internet Myths”

  1. the “separation of distribution and content” means that you no longer need to be the network owner to deliver the service; maybe it could have been more clearly stated as “the application being separated from the transport and, thanks to full interoperability (implemented, not theorerical), even provided by operators on other networks”

    this independence from the network operator [and interoperability and peering agreements] allows to deliver any service worldwide, a significant opportunity for anyone who can harness it, yest, if services are blocked, the paradigm fails.

    I think that stating that the internet “does not differ in terms of structure” is so general that one can can find different levels at which it is obviously true. After all, it’s all wires and boxes, isn’t it ?

    it all depends on the level of abstraction of the discussion.

    Would you say that theactual implementations of 3GPP have the same structure of the implementations of IP ? Network access control by operators is such that any application on a 3GPP network needs to be approved and deployed by an operator; it seems to me this is a major structural difference. I could argue the same about X25, Minitel, Q921, etc.

    the internet differs from old technologies in many, many ways and the independence of the service from the network layer and the interoperability and peering agreements on the Internet allowing to seamlessly deliver services worldwide, IMHO, are two major structural differences.

  2. Stefano, you’re falling into the same trap as Wu. In the days of the original analog telephone network I could deliver any service involving voice — from the time to stock quotes to consulting to dial-a-porn — over the telephone. No different from the Internet of today.

    It’s also worth noting that the network management practices which the rabid “network neutrality” lobbyists are lobbying against (and which was addressed badly in the Comcast proceeding at the FCC) was simply intended to guard against behavior which harmed the network or hogged its resources. That network management was content-neutral. It didn’t care whether the content was kiddie porn or the King James Bible. So, your arguments above do not apply.

  3. Richard said:

    “This is a big contrast to the telephone network, where information is constrained by call setup delays and a very narrow pipe.”

    Narrow pipe? It’s a fixed sized pipe if you’re talking traditional TDM 🙂

    Voice delivery over IP networks and wireless doesn’t do this necessarily; although the separation of signaling data (SIP, MGCP) and the call stream (RTP) is analogous to the current voice network.

    With modern carrier optics, the distinction between Traditional TDM and Cell/packet based forwarding is usually virtualized at the physical wire (DWDM) ; or converted from one to the other when it hits the edge of the network (hence the use of media gateways.)

    Everyone in this game is using the same stuff for back haul and Interoffice/Headend transport; the only difference between “telco” and “cable” from a technical residential perspective are the wires coming into the house. I understand it used to be much different, but not anymore.

    Seems to me that’s a pretty thin line to base regulations on; and the government is doing a huge disservice to public by keeping up this charade.

  4. The problem with not “keeping up this charade” is that regulators and lobbyists will opt to push everything to the side of the line that gives them more power: that is, more regulation rather than less. And if all telephony (even VoIP) becomes subject to what’s called “Title II regulation,” which used to apply to the telcos and is biased toward them, the public will suffer a great deal.

  5. HI Brett,

    “The problem with not “keeping up this charade” is that regulators and lobbyists will opt to push everything to the side of the line that gives them more power: that is, more regulation rather than less.”

    This is exactly what pisses me off; although I’d be in favor of qualitative based regulation than quantitative; but that’s a pipe dream under today’s legal structure; engineers would need to run it 🙂

    In regards to Title II, the only benefit Telco’s have is that they’ve already spent gobs of $$ on compliance and integration; I don’t necessarily agree with the proposition that asking for compliance from similar service providers will impact consumers in regards to pricing vs. Cable (Basically, it’s $100 a month minimum for triple play anywhere cable/fios/whatever ) but it will be a significant PITA for the independent IP dial tone providers like Vonage.

    It should not be a quixotic quest to take on the “talking through the handset” based regulatory framework; since alternative means of meeting regulatory requirements for things like “emergency services” and “universal access” are not that difficult to achieve given the proper incentives; being subject to the whims of the flaky and inconsistent FCC while more enlightened countries roll their eyes is a barrier to investment and it hinders progress.

    However, I agree with you in the sense that I have very little confidence that the correct solution to this problem is the creation of “Teenage Bells” like Verizon and AT&T.

    If the government wanted to get involved, I would have a single mandate.

    1) Buy existing last mile networks at a fair price. (Since we’re failingout Wall St. and Detrioit, why not?)

    2) Fund and promote the development of last Mile Ethernet Networks to neighbor hoods and homes; it can be delivered over copper or fiber, and is the most cost effective interconnect (sans wireless, but that is out of scope for this) to deliver services to customers.

    It’s open, it’s an actively developed IEEE standard, and both cables and telcos use it after the first non subtended hop to deliver services due to it’s ubiquity. I believe it would provide the most benefit to end users in the long run. It’s basically squashed everything else in the Enterprise and backhaul space (RIP ATM, Broken Ring etc.); so it should be a low on the controversy scale.

    Regulate based on the layer 2 / Layer 3 hand offs, by this I mean allow the end user or via LOA their L3 provider to control the mapping mapping of CoS to DSCP for their applications; with exceptions for “emergency” communications.

  6. Would it really be a good idea for the government to buy me out? Could it even do so, given that my ISP is not a public corporation (there’s no stock to buy)? And would I want to work for the government?

    I don’t think that nationalizing the Internet is the answer.

  7. HI Brett,

    “Would it really be a good idea for the government to buy me out? Could it even do so, given that my ISP is not a public corporation (there’s no stock to buy)? And would I want to work for the government?”

    You misunderstand my point; I merely suggested that the Government buy out the current physical last mile networks to allow $ENTITY (Muni, Internet exchange style co-op, Private company) to deploy a pervasive metro ethernet fabric; not to Nationalize “the Internet”. Things properly classified as an ISP would be a Layer 3 provider, and would remain relatively unregulated; seeing as we don’t really have issues with long haul capacity.

    What I’m suggesting gutting the current Telecom/Cable regulatory cognitive dissonance and making the entire point moot by transforming wire based (in a residential, regulated sense) last mile providers into entities that operate like Switch and Data and other Internet Exchange operators.

    This would make hooking up with your customers easier than dealing with LECs and the FCC.

  8. Max:

    As a wireless ISP, I operate a full fledged, infrastructure based, physical “last mile” network that is every bit as good at delivering broadband than copper, fiber, or coax. And it is absolutely part of the Internet. Would you have the government appropriate my business? Or engage in unfair competition with me (since, after all, it has taxpayer subsidies and infinitely deep pockets) and destroy my business and that of my 4,000+ colleagues?

    What’s more, you are very incorrect when you say that “we do not have problems with long haul capacity.” Here in Laramie, Wyoming, it is impossible to get backbone bandwidth for less than $100 per Mbps per month. I have been attempting to work around this, but have been effectively stonewalled by the national fiber backbone providers (whose fiber passes through the Laramie Valley without stopping). They could open up their pipes to us, but they will not at any reasonable price. The best offer we’ve gotten, when equipment amortization and startup fees are factored in, is close to the $100 per Mbps we are already paying. In short, long haul transport is a big and very serious issue. Congress and the FCC have put it on the shelf under the term “special access” — a misnomer because in fact there is nothing special about it. It is the general, necessary wholesale access which is required to offer Internet service.

  9. HI Brett,

    “As a wireless ISP, I operate a full fledged, infrastructure based, physical “last mile” network that is every bit as good at delivering broadband than copper, fiber, or coax.”

    Wireless is fantastic, don’t get me wrong, but such transmissions have plenty of regulations already vis a vis spectrum. As I said earlier, the problem that needs attention is physical wireline facilities.

    “What’s more, you are very incorrect when you say that “we do not have problems with long haul capacity.” Here in Laramie, Wyoming, it is impossible to get backbone bandwidth for less than $100 per Mbps per month.”

    $100 per Mbps per month in *fees*? If that was the case, you’re business is likely underwater unless you are you also including your overhead in terms of colocation plant; but you’re going to need to pay for that anyway; whether the equipment faces the edge or the core of the network.

    And you’re correct, there is nothing special about it. That’s the point!
    Sure, you need to buy or lease fiber, purchase transport equipment; make sure that your gear is compatible with the existing ring and build out to a suitable splice enclosure. Compared to rolling out the same thing to 10,000 homes, handling backhaul is so cheap that anyone can (and does) do it (Cogent); and it’s a much more competitive playing field; you usually have 4 or 5 choices vs. the duopoly of LEC or cable plant you have in terms of last mile connectivity.

    I’m a bit ignorant of the specific geography in your locale, but I’m assuming your close to a university, which means there’s a carrier hotel in 200 Mile Radius from your POPs. If you can build to those locations, wholesale access[1] shouldn’t be an issue.

    In absolute terms, sure it’s expensive, but you’re the ISP, not the end user.

    [1] We didn’t define this specifically, but I’m guessing it’s either wavelength or ethernet cross connects for IP peering/transit you’re referring to.

  10. “Would you have the government appropriate my business? Or engage in unfair competition with me (since, after all, it has taxpayer subsidies and infinitely deep pockets) and destroy my business and that of my 4,000+ colleagues?”

    I forgot to touch on this. But under my plan there is no current issue with wireless providers (esp. unlicensed spectrum). You don’t own tonnes of rotting copper and barbed wire transmission facilities. In fact, ISP’s ilke yours are a work around and potential solution to the very narrow problem I’m focusing on.

    Again, I want to the Last Mile Physical Wireline network to ideally be rebuilt and operated in a manner similar to the neutral facilities provided at Internet Exchanges.

    Independent service providers (like you) would be charged neutral wholesale fees to access this network via their own facilities (Their existing metro and longhaul).

    The “Exchange” operator basically has only one set of customers; the service providers; so it removes the existing conflict of interest that LECs and Cable cos. when it comes to their role as Internet Service Providers. Thusly, you could use these facilities to provide “local” backhaul to carrier hotel hub sites; or you could “build your own” and do something analogous to private IP peering.

    This also puts the focus on a regulatory framework that is data transmission focused, instead of hairsplitting over the outdated regulatory focus that seems to be based on whether your primary business is voice delivery or video delivery; with data as a cutesy add on.

    Philosophically, the reason I choose the Internet Exchange Model is because it has a good history of regulating it self just fine without a lot of government involvement and requires cooperation between all parties in order to function properly. That combined with a technological bias towards Ethernet (a standard that also as evolved without a lot of government intervention; it also has the added benefit of being a commodity ) should make just about everyone happy.

    Frankly, it’s more a less just a matter of time before this happens anyway; I’m merely suggesting that we give it a big shove in the current climate of throwing dollars at major infrastructure upgrades and failing industries. This does seem a bit more sensible than throwing dollars at lawyers to gain competitive advantage via FCC lobbying; assuming everyone has the end user / consumer/ subscribers best interests at heart.

  11. Max, you have so many misconceptions that I don’t know where to BEGIN to set you straight.There isn’t any competition here for backbone bandwidth. One company owns all three of the backbones that pass through our area and will not open them up. And the telephone and cable monopolies want to charge us more at wholesale than they charge their end users at retail. There is ZERO competition. And building fiber out at $150K a mile isn’t even an option; we are 45 miles from the next city over.

    And as for the last mile: you’re now talking about building out redundant infrastructure to compete with me at government expense. And then possibly trying to wholesale it back to me, which does no good because at best I get the same deal as everyone else and cannot differentiate my product. And at worst I get a worse deal because I’m small or am excluded because the network gives one provider exclusive access. Sorry, no dice. Don’t try to build my network for me. Just get out of the way and let me build mine.

  12. “There isn’t any competition here for backbone bandwidth.”

    WTF?

    I could get an OC-3 coast to coast in 2000 for less than $9k a month; to put things in perspective.

    I understand that this is seriously getting off topic and that Richard might kill this off just to avoid the noise, but I did a quick google search for bandwidth in Laramie Wy and I was getting *retail* quotes back that were lower than what I was paying back in ’00.

    “And as for the last mile: you’re now talking about building out redundant infrastructure to compete with me at government expense.”

    I paid for the stuff you’re doing now at government expense; if you consider wireless spectrum politics 😉

    You differentiate your product by offering better service; and interesting features. I’m sorry, but “unique” CoS to DSCP mappings don’t cut it; and fixing busted cable is mostly an outsourced practice at this point.

    What is ironic is that you are complaining about expensive transport and backhaul on the one hand, yet you don’t seem to grasp basic carrier scale economics on the other and how the solution I’m proposing addresses the issue you’re complaining about. That and, unfortunately, fighting strawmen; I clearly indicated the preference for minimum government involvement in this process. I believe there is universal consensus in the idea that Chris Dodd et. al. should not be dictating network management policy to anyone.

    For the record, don’t bother trying to set the record straight with me. I’ve participated in the development, construction, and operation of very dense metro networks and Coast to Coast long haul networks for major carriers; and have experience with the issues for both wireless and wireline ISPs. If you want to challenge my proposal, please do it on the technical merits. I’m certain that there are plenty of problems with it that you can address on that level.

    I would like to hear them.

    ~Max

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.