February 25, 2008
[fccboston08] FCC Hearing: Panel 2: Q&A
NOTE: I am live-blogging. Not re-reading for errors. There are guaranteed to be errors of substance, stand point and detail. Caveat reader. |
McDowell: Does BitTorrent run differently on different types of network infrastructures?
BT: The engineers do take advantage of architectures. We rely on the basic structure of the Internet to handle congestion controls. The Internet has been doing that well.
McDowell: Is that what BitTorrent DNA does?
BT: It’s a competitive market, so we’ve worked on theability to run well in particular network envrionments.
McDowell: Should you be required to disclose that you don’t work well in particular types of networks?
BT: BT emerged as an open source protocol. We run it like an open standards process. We disclose it all. As issues materialize on particular networks, there’s demand for solutions, and we see a very open process.
Weitzner: The premise on the Internet is that there are standards that work, and the nodes and apps conform to those standards. If Comcast markets as Internet service, it ought to support Internet standards. Apps should be able to assume that they are operating in an Internet-compliant environment. Apps can’t conform when the carriers aren’t conforming and when they’re not given info.
Clark: In the beginning, we thought the apps should adapt to the underlying network. The user ought to decide what’s satisfactory not the app. E.g., a slow file transfer might seem unusable, but users find uses for it. In the end, it’s the user who should decide to hit the quit button. But crossing into paying for a service and expecting to get it requires adapting the net and the apps. We need tools for talking cross layer. There’s space for technical innovation.
Reed: Adapting to categories of networks traditionally have been the role of the IETF. E.g., 20 yrs ago there was a move to asymmetric networks. The community wanted to ban it. More pragmatic folks in the IETF said that it’s a reasonable technology an we have to be able to live with it. But that was the standard Internet, not the Comcast subclass of the Internet. If there are changes through the IETF, networks and apps will adapt. But it’s different when one vendor has a lock on one area and makes unilateral changes.
Now we see videos made by citizens today.
A retired fed lawyer says she needs the Net for her work playing music for cancer patients.
A worker for a small non-profit that makes a small TV app for participatory culture. Comcast’s filtering of BT harms his company and anyone who tries to post without going through a central authority. Comcast is “essentially ruining our system.”
A MIT computer science student sees it as a political issue. What will happen to MIT when it begins writing papers about the negative effects of this type of censorship. It affects the free flow of info.
The Maine Civil Liberties Union likes NN on First Amendment grounds.
Martin: David Clark, you want to change the business model…
Clark: Charging by the byte works poorly. But it might be worth exploring usage caps measured not in peaks but in gigabytes per month. [So, not $50/month for a max of 10Mb at any one time, but $50/month for a total of 5Gb (or whatever) of downloads.] The wireless guys are breaking the ice here. The current all-you-can-eat model has the advantage of encouraging people to try things, but it hides the cost and the value.
Martin: That could make sense in terms of business and maybe address some of the congestion problem. But with BT, it was their uploads that were being blocked, and they were below their limits anyway.
Clark: The ISPs haven’t characterized what acceptable use is. This is understandable because it’s very hard to know exactly what the conditions are going to be. But we can determine fairness of use among users without looking at the particular applications being used.
Martin: Do you think it’s important to distinguish among types of bits, not applications?
Clark: If there’s going to be any discrimination based on QoS, I’d rather have the user determine it. “This telephone call is really important!” Maybe 10% of your bits could be high priority, and you decide.
Martin: Was BitTorrent designed to get around some of the limitations of networks such as the cable network?
BT: The protocol was designed simply to manage large files. It doesn’t behave differently on different types of networks.
Reed explains a bit about how the IETF works. Locking in decisions around apps and QoS stops innovation in its tracks.
Bennett: It takes and months to get a solution through a standards body. Comcast has an obligation to solve it faster and to do the experiments.
Copps: Do we have enough info to proceed?
Clark: From the perspective of the academic the answer is simple: No.
Reed: My colleagues who try to collect data report a systematic shutting down of info.
Copps: Is this a serious problem?
Clark: Yes. E.g., we’re trying to find out if 10% of users generate 90% of the traffic. The only info I could get was from s. Korea. Not a single US ISP would give us the info even fully laundered. We don’t even know if there are consistently heavy users. They say it’s proprietary.
Copps: we really need to grapple with this.
Martin: Does that mean we shouldn’t take action, because we don’t have enough info?
Reed: No one has enough data to know whatt the current or projected traffic across the entire industry is.
Clark: The complaint before you does not require detailed traffic knowledge to make a decision.
Adelstein: How important network mgt practices for our ability to maintain our leadership in innovation?
BT: Incredibly important.
Reed: I spoke with the head of China’s largest portal. He’s got the rights to distribute the Olympics over their Net. That’s in competition with the state TV service. They may pass us with this new tech.
Bennett: If the FCC gets the decision right, you’ll increase the value of the carrier’s infrastructure. If you take the carrier out of it, the blocking will happen but it will be done by the guy down the street. The carrier is acting as a sheriff. If you add enough capacity so that there is no congestion, you’ll have to increase it 10- to100 times.
Clark: I disagree. The network does contain mechanisms to control traffic. The Comcast is a particular, nuanced response to what the Internet’s been doing for years, which is when too much traffic shows up everyone slows down. Being slowed down by our neighbor is right and it’s a good thing, because it means that when my neighbor isn’t there, I can go twice as fast. The question is whether the net is allocating in a way the user thinks is fair. (Should you be able to buy your way out, he asks) If you want to have nuances, you have to be clear about it or people won’t bother to innovate. No one wants to compete with Vonage because the carriers have done certain things.
Sony: Short term solutions should be recognized as having global implications.
Weitzner: The video innovation we’ve seen today is just the beginning. We ought to be working to get past the relatively mundane question of how to transport video in a civilized way.
Clark: There’s a world of innovation waiting to happen when there’s enough bandwidth.
Adelstein: Where is the line between sound network mgt and unsound? Between good discrimination an bad?
BT: I know it when I see it. Comcast is doing what [bac] hackers do to disrupt traffic. Forging data is outside the realm of reasonable.
Bennett: What I found lacking in the petitions against Comcast was any data about the condition of the network when the mgt practices were put into effect. They didn’t provide empirical data about how busy the Net was.
Clark: Using RST indicates that Comcast’s tools were inadequate for predicting and managing.
Bennett: The data is available.
Reed: I checked it out myself. They were indeed sending RST packets. They have to be carefully synthesized. And I discovered that to do that they had to use info from inside packet. I would view that as a pragmatic solution to the problem if someone had analyzed this technique and published a paper that says RST is a good way to solve congestion. But I can’t find anything. There is a bright line around hacking without public data.