About Tom Evslin

Video Profile of Tom Evslin

Follow Tom Evslin on Twitter


subscribe:

Add to Technorati Favorites!
Powered by TypePad
Member since 01/2005

technorati


« Interview with Me | Main | Look Into the Good News »

Reader Comment on Net Neutrality (and My Rebuttal)

Reader Richard Bennett, whose own blog often articulates the telco point of view on Internet Neutrality more coherently than most telcos, posted a challenging comment to my post announcing that I’m speaking at the Berkman Center tomorrow.  His questions deserve answers which I’ll give in this post and also tomorrow when I speak.  Richard’s whole comment is quoted below but I’ve taken the liberty of interspersing my answers.

Richard: “You say: I very much believe we need an Internet which remains essentially neutral and application agnostic.

“Perhaps you could explain why you believe that and what it means to you. For example,

“* Do you believe that all applications have the same requirements from the network?”

Answer: The reason we need an Internet which is network application agnostic is because only an application-unaware network allows the kind of incredible innovation which has occurred on the Internet.  David Isenberg (who Richard mislabels on his blog as a marketing person – David’s background is Bell Labs) explains why this is so brilliantly in The Rise of The Stupid Network).

My simplistic explanation is that a network which is built to be aware of specific applications may work well for those applications that the network designers had in mind but will inhibit if not absolutely prevent applications the network designers didn’t think of as well as the evolution of existing applications.  The traditional phone network is a great example of that.  Handsets on that network don’t even have the capability of cell phones because the network is optimized for voice and voice alone.  I posted more on this here.

To answer the starred question: I KNOW that different applications rely on the network in different ways.  We get to the implications of this in Richard’s next question.

Richard: “* Do you believe that applications that require tons of bandwidth, such as movie downloads, should have the same service from the network in terms of average packet delay as applications such as VoIP with modest bandwidth requirements but more stringent jitter requirements? (note: "jitter" is the variation in packet delay.)”

Answer: Yes, I do.  I do know what jitter is since I founded and ran the world’s largest wholesale carrier of international VoIP and am a nerd but I do appreciate the explanation.  Richard is right to say that VoIP has modest bandwidth requirements and is sensitive to jitter (because the human ear is very sensitive to breaks in speech).  In fact, this is a wonderful example of both why a neutral network is needed and why it succeeds.

When Mary and I founded ITXC in 1997, the common wisdom was that voice over the public Internet couldn’t work because the Internet is a chaotic place and does not promise quality of service in any dimension including packet delivery times, jitter, or even reliable delivery of packets.  In fact, Internet routers throw packets away when they are overburdened.  Moreover, the Internet certainly wasn’t designed with voice in mind. But it IS application agnostic.

Guess what?  We found that we could do exactly what David Isenberg said: put intelligence at the edge of the network – actually the interface between the public switched telephone network (PSTN) and the Internet.  Some of this intelligence – jitter buffers for handling jitter, for example – was designed by gateway vendors like pioneering VocalTec and later Cisco and others.  Part of it was home grown: technologies for discovering which routes over the Internet were currently giving us the best quality and value and using them.

It worked.  When ITXC was sold in 2004, it was the seventh largest carrier of international minutes in the world (measured in minutes carried) – this includes all the traditional carriers.  Note that virtually all of these minutes both originated and terminated on ordinary phones.  The callers did not know that their carriers – most of the major carriers of the world and many new ones – were handing the calls off to us for delivery to the called-country via the Internet where our affiliates then fed them back into the phone network.  We would not have been successful if we could not deliver the quality of service which these callers and their carriers expected.

The point is that we succeeded BECAUSE the Internet didn’t try to help us.  We were able to build what we needed at the edge of an application-agnostic but open network.  File sharing services, video distributors, and other Internet services all do their own customization of the network AT THE EDGE.  That’s why the network does NOT have to try to figure out what services each application needs and why revolutionary new applications can be developed.

There would be no VoIP today if we were waiting for the owners of Internet pipes to optimize them for our application.

Richard: “* Do you believe that it would be a good thing for Americans to have fiber-optic tubes running between their homes and the public Internet? If so, do you believe that the companies who install the tubes should be allowed to make a profit from their operation?”

Answer:  That’s easy.  Yes and Yes.  More bandwidth costs more.  I would happily pay more for more bandwidth if it were available in Vermont.

Richard: “* Do you believe that it's good for content producers to profit from the advertising they move down the tubes, but no one else?”

Answer: I don’t accept the premise that only the content producers benefit from the advertising.  Internet users benefit because they receive the content which is paid for by the advertising.  The advertisers  benefit or they wouldn’t advertise.  And the Internet Service providers benefit because they are paid for faster connections both by consumers who want the content and by the content providers. We wouldn’t need megabits of speed if we only did email.  In fact, in a recent deal, Verizon agreed to pay for content in order to make their newest network more attractive to the end users who pay a premium (properly) to connect to it.

Richard: “* Do you believe that triple play is legitimate or illegitimate? And does your answer depend on whether triple play segregates traffic by frequency or by time?”

Answer: By triple play, I believe Richard means the offering of phone service, Internet access, and entertainment in a bundle.  Yes, I do think it is legitimate for one company to offer all three services no matter how they’re delivered technically. As a technician, I believe it’s a bad idea to segregate by frequency rather than time on a new network (the Internet segregates packets and applications by time) but that doesn’t mean I’d outlaw frequency segregation.  Both DSL and today’s cable-provided Internet access are examples of segregating by frequency which is OK as a transition technology.

What I object to is the ability of the telco and cableco duopoly in the US to use their duopoly position to advantage their own phone and entertainment services when they are also the underlying provider of capability.  This is exactly analogous to the dominant provider of operating system software using the operating system to advantage its own applications.  If there were a competitive market for Internet access in the US,  I would think that discriminating would be bad business for an access provider and that the market would sort that out – assuming that discrimination was known which is what I’m speaking about at Berkman.

Richard: “* Do you believe that companies who invest in infrastructure are entitle (sic) to recoup their investment in part by selling services on the infrastructure that are advantaged technically over services sold on the same infrastructure by companies who have no infrastructure investment at risk?”

Answer: Yes.  But, if the companies that you’re talking about are part of a duopoly, then I believe that they must make it equally possible for their competitors to technically advantage their applications whether they own infrastructure or not.  The advantage can’t come from duopoly control of the infrastructure.  See the Windows analogy above.  Microsoft has to make its APIs available to application developers who have NOT invested in the infrastructure of an application.

Richard: “* And finally, do you believe meaningful measurements can be reported by people who don't understand what they're measuring or why?”

Answer: Yes.  So long as they have the right software.  That’s exactly what I think and what I’m recommending.  Note I have NOT called from Net Neutrality regulation or legislation despite my fears of the duopoly.  I am calling for information so that the marketplace and, if necessary as a last resort, the political body can act.

Richard: “I won't be listening to your talk at Berkman, but will most likely check whatever podcast is available after the fact.”

Answer:  Good.  Thanks for helping me plan the talk (sincerely) and I look forward to your comments after.  This is an issue which needs more light and less heat.

TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d83451cce569e200d8343e4b0753ef

Listed below are links to weblogs that reference Reader Comment on Net Neutrality (and My Rebuttal):

» Tom Evslin answers my questions from The Original Blog
I left some questions for Tom Evslin on the talk he gave today at Berkman on neutrality regulations. Tom has a technical background, and while hes not come out in favor of new regulations, he appears sympathetic to the arguments for them. Check... [Read More]

» My Rebuttal (?) to some Comments on Net Neutrality from Aswath Weblog
This morning Tom Evslin posted his responses to comments/questions raised by Richard Bennett, one of his readers on the topic of Net Neutrality. According to Tom, Richard “often articulates the telco point of view on Internet Neutrality more coherently... [Read More]

» My Rebuttal (?) to some Comments on Net Neutrality from Aswath Weblog
This morning Tom Evslin posted his responses to comments/questions raised by Richard Bennett, one of his readers on the topic of Net Neutrality. According to Tom, Richard “often articulates the telco point of view on Internet Neutrality more coherently... [Read More]

Comments

sohbet

The alternative to a single-purpose network is a multi-purpose network that can recognize the needs of applications either directly - because it's told what they are - or indirectly - by observing traffic, but that's only for legacy applications that don't know how to request a service level from the network. And we've just begun to learn how to build multi-service networks since 1990 or so.

Your voice service altered the architecture of the end-to-end public Internet, didn't it? Nowhere in the RFCs or in the practice is there a mechanisms whereby applications can discover routes and pick and choose among them. By embedding this function in the network, you violated Isenberg's "dumb network" and gave it the kind of intelligence it needs to have for its multi-purpose future.

In other words, you found a way to wring priority service out of a dumb network, and it worked largely because nobody else was doing it. So it seems to me that the success of VoIP with ITXC is an example of how application innovation works best on networks with multiple services, not just a single, one-size-fits-all model.

And it's interesting to note also that what you did with ITXC would be illegal under the Snowe-Dorgan and Markey regulations because you implicitly provided enhanced Quality of Service for a fee to voice users. Good thing you don't support the regulations.

Richard Bennett

Listening to your podcast, I noticed that some wag felt compelled to interrupt you with a declaration that I'm a notorious troll, not an engineer, and not a serious "seeker". While I'll cop to the last one, the others are essentially slanderous. I'm co-author of the IEEE 802.3, 802.11, and 802.15.3a standards and related items.

I'd like that person's name, if you don't mind.

Richard Bennett

OK, Aswath, now we have two theories about how IXTC worked. But in any case, what does this case tell us about "net neutrality?" Does the consumer have access to multiple NSPs concurrently, with the ability to control the switching of packets between them in real-time?

I don't think so. And this is one of my criticisms of the argument that the Internet is a Stupid Network that runs totally under end-user control. If that formulation had a kernel of truth to it, verily the user would be able to select his own routes depending on his own criteria. But the Internet is in fact a smart network that handles its interior queues as it wishes and selects routes as it wishes.

The Telco plans to offer tiered service actually do place more of these things under the control of the end-user, and that terrifies the established business interests who've figured out how to exploit the present system's flaws. Put yourself in Google's position: you've invested hundreds of millions of dollars, placing 450,000 computers at 25 points on the Internet. You can out-perform any startup capitalized for less than $500 million.

Do you want this system to change?

Aswath

Richard:

I think you are mischaracterizing ITXC's route selection. My reading of Tom's description suggests that ITXC gateways had different access providers, they had a mechanism to determine which access provider's path had better performance and the gateway selected that access provider. This can happen even when the individual access providers conform to Net Neutrality. For example it so happens that at that time not many customers of an access provider is not generating traffic. In this case, ITXC is an "end-user". So I am not sure that this form of "route-selection" is in violation of these acts - in letter or in spirit.

Richard Bennett

John B, Tom says the ITXC product used proprietary algorithms to find routes with high QoS. This is not something that the end-user interface to the Internet allows, and it violates the idea that all the Internet should ever do is offer "best effort" delivery. Obviously, some efforts are better than others, and for a price.

If a last-mile provider offered an ITXC-level service to its end-users, and Snowe-Dorgan was the law, then the ISP would be in violation of "net neutrality." That's why Snowe-Dorgan's supporters are quick to label ITXC as an application instead of the infrastructure enhancement that it clearly is. We can quibble about how the ITXC service has to be bundled in order to escape jurisdiction of Snowe-Dorgan, but it's abundantly clear that if it were under that jurisdiction, it would be illegal.

I think it would be really swinging if end-users of the Internet could select routes according to the type of traffic stream they want to move, and that's the entire reason I'm against the Snowe-Dorgan and Markey regulations.

Edward

I appreciate the two of you engaging in this debate - it's been far more intellectually honest than what we're hearing in other more public forums.

Incidentally, and on that point, I ran across this article (http://www.telecomweb.com/tnd/18535.html) in TelecomWeb suggesting that the top brass at Comcast and eBay are engaged in a bit of a back-and-forth that seems to echo Tom's call for "more light and less heat."

Also glad that you both specifically addressed jitter and other traffic management realities - this was a key focus of Ed Felten's paper (http://itpolicy.princeton.edu/pub/neutrality.pdf) on this issue and I felt it didn't get much play beyond his point-counterpoint with Bill Herman at Public Knowledge (http://www.publicknowledge.org/node/523). I happen to think that's a far more interesting topic of debate than, say, whether or not individual bloggers will have to "pay-to-play," with I think mischaracterizes the debate.

I suppose that's a long way of saying "thanks" ...

John B

Richard,

I think your rebuttal is deeply flawed, in a couple of areas (and, for my part, I have a MSc in CompSci w/a focus on distributed networking, and I worked for Nortel on voice switching software for many years).

a) ITXC did not alter the architecture of the public Internet. It did not embed the function in the network. It used the existing standard routing protocols, the existing standard best-effort UDP. The application code at either end were designed to compensate for the weaknesses of UDP. That is not, by any reasonable stretch of the imagination, the same thing as what SBC wants to do.

b) I can't find the text of Snowe-Dorgan, but as far as I can tell, it regulated network operators. Was ITXC a network operator? (Were they an ISP or NSP) - or just an user of existing networks?

c) SBC and Comcast, et al, have monopoly license to build their plant and lay their coax and fiber and twisted pair. I am a rabid free-marketeer, but even I recognize the tremendous strategic business advantage the telephone and cable companies have developed by acquiring that monopoly license. And in the end, they were given those licenses because it would facilitate the rapid deployment of infrastructure, which was in the public benefit. Having accomplished that, I don't see how being able to reduce their monopoly customer's quality of experience is in the public benefit.

d) Imagine you are having a house built, and the electrician comes and wires it up. Afterwards, you plug in your appliances, and find that some don't work. You call up the electrician and he says "Oh, yeah, I have a deal - only GE appliances get full amperage through your house network. Everyone else gets 1/2 the normal amps. How else do you expect me to afford that Porsche?" Wouldn't you be livid? I know I would be.

Bottom line: The phone and cable companies get a huge amount of money every year because they have or had monopoly control over the wire, a infrastructural benefit that will last for dozens, if not hundreds of years. They don't deserve to get any more by extorting the people (and the descendents of the people) who gave them that monopoly position.

Richard Bennett

So how'd the talk go? I'd guess you must have got a lot of "Hallelujahs!" with your "application-agnostic-spurs-innovation" line.

Tom Evslin

Richard:

You are right that I meant "application agnostic". I fixed that. Thanks.

ChadB:

You are right. ITXC was not a broadband provider by any definition. Thanks to you also.

Kempton:

Thanks for you encouragement.

ChadB

Richard, the Snowe-Dorgan amendment applies only to "broadband service providers," which is limited to "a person or entity that controls, operates, or resells and controls any facility used to provide broadband service to the public." Would that apply to ITXC?

Kempton

First of all, what a pleasure to read the exchange between Tom and Richard. Your discussions here have definitely helped me to become a lot more informed on the relevant facts and debate on this very important issue.

Tom, I really look forward to the webcast of your talk at Harvard Law School later today.

Cheers,
Kempton
Calgary, Canada

Richard Bennett

Thanks for the response, Tom, I appreciate your desire to shed some light on this debate, which up to this point has been mainly people talking past each other, bumper-sticker slogans, and fear-mongering. Whatever action Congress decides to take in the question this year - and it may be no action at all - the people benefit only if they act out of knowledge rather that emotion.

Let's take the first question and drill down a little deeper, and if that takes us anywhere interesting we can address the others in turn.

It seems to me that you're conflating two similar but distinct concepts in your statement about application-agnostic (you said network agnostic, but I assume that was a typo) and application-unaware networks. The Internet, like all networks designed in the 70s, isn't application agnostic even though it may be application-unaware (it's not that either, but we'll get to that later.) The Internet was designed to serve the needs of file-transfer applications such as e-mail and ftp, and not the needs of real-time applications such as remote sensors, gaming, and voice. It's essentially a store-and-forward network by design.

And for good reason: its architects didn't want to duplicate the telephone network, they wanted a testbed for learning about packet-switching, and that seemed to be mainly a bulk data transfer enabler to their naive (we were all naive about packet switching in the 70s) minds.

A network that is designed around the needs of one class of applications doesn't need to police traffic to ensure that only its cardinal application gets proper service, it has a bias in its design that makes it act that way. So while the Internet has given rise to a number of innovative applications such as web surfing, auctioning, and social hookups, all them are of the same type from the network's point of view: they do things that look an awful lot like file transfer.

The alternative to a single-purpose network is a multi-purpose network that can recognize the needs of applications either directly - because it's told what they are - or indirectly - by observing traffic, but that's only for legacy applications that don't know how to request a service level from the network. And we've just begun to learn how to build multi-service networks since 1990 or so.

Your voice service altered the architecture of the end-to-end public Internet, didn't it? Nowhere in the RFCs or in the practice is there a mechanisms whereby applications can discover routes and pick and choose among them. By embedding this function in the network, you violated Isenberg's "dumb network" and gave it the kind of intelligence it needs to have for its multi-purpose future.

In other words, you found a way to wring priority service out of a dumb network, and it worked largely because nobody else was doing it. So it seems to me that the success of VoIP with ITXC is an example of how application innovation works best on networks with multiple services, not just a single, one-size-fits-all model.

And it's interesting to note also that what you did with ITXC would be illegal under the Snowe-Dorgan and Markey regulations because you implicitly provided enhanced Quality of Service for a fee to voice users. Good thing you don't support the regulations.

Post a comment

If you have a TypeKey or TypePad account, please Sign In.

Now on Kindle!

hackoff.com: An historic murder mystery set in the Internet bubble and rubble

CEO Tom Evslin's insider account of the Internet bubble and its aftermath. "This novel is a surveillance video of the seeds of the current economic collapse."

The Interpreter's Tale

Hacker Dom Montain is in Barcelona in Evslin's Kindle-edition long short story. Why? and why are the pickpockets stealing mobile phones?

Need A Kindle?

Kindle: Amazon's Wireless Reading Device

Not quite as good as a real book IMHO but a lot lighter than a trip worth of books. Also better than a cell phone for mobile web access - and that's free!

Recent Reads - Click title to order from Amazon


Google

  • adlinks
  • adsense