« Scenes from the UN General Assembly | Main | The Sound and the Fury »

September 24, 2009

Comments

I'm a researcher in EE who has been following these discussions (along with the more technical discussions of differential pricing that go on in the academy) and am shocked at the base ignorance (or hypocrisy) of the popular writing on the subject.

The conflation of the application layer with QoS may lead (theoretically) to some performance "improvements," but the fact of the matter is that it will let providers who have benefitted from government largess to dictate the new application developments and extort money from application developers.

First, Reason is no better than CATO, which means good on civil liberties, and so-so to batsh*t crazy on economics (part of the Chicago meltdown?).

Second, notice that Suderman, in the midst of bashing government, uses FCC measures known to be really, really lousy, to justify bashing government. It's like declaring that an opponent of Stalin *must* be guilty of lies in accusing Stalin of oppression, because the guy confessed at his trial.

So that said, there is a great deal wrong with this Peter Suderman ...

This is about there point where I stopped reading. Does it make me a bad person if I see the author's name, know I'm going to confronted with libertarian sophistry and disingenuous claptrap, and ignore it?

Have you communicated your critique to Mr. Suderman, publius?

It bothers me (somewhat) that 'libertarian' has come to mean 'social-Darwinism on the make'. I agree w/ JustMe's sentiment, but all the so-called 'libertarian" arguments I hear anymore are one-legged at best...and sophistry & clap-trap to boot.

Suderman going all Emily Litella over "government regulation of the inner-tubes" (ie, net neutrality) is very much a case in point.

I think Publius has the right handle on this issue, and that JustMe isn't a bad person simply because s/he chooses to bypass another source of empty mental calories...

publius--

Honestly, I just don't follow this one. I think maybe he's not understanding the idea of network layers. The FCC's actions are primarily about protecting the "bottom" physical layer -- it has nothing whatsoever to do with higher-level application layer services where innovation lives.

If you understand the concept of layers, you're understanding it in an old, bell-headed way, rather than in the way that internet engineers think about it. For example, if the FCC were to prohibit things like the Comcast BitTorrent throttling, they'd have to be regulating at the transport layer. That's layer 4 in most network models (1=physical, 2=datalink, 3=network).

This is pretty much where the ISP's purview ends and that of the application or content provider begins. That's a defensible place to draw the line, but your formulation makes it sound like it's not a big deal. It's a big deal for the ISP.

You seem to be worried about protecting consumers and content providers from their ISPs, and, to a certain extent, protecting smaller content providers from larger content providers. No problem there. Do you have equal concern for protecting ISPs from consumers and content providers? Would you want the FCC involved in regulating that sort of protection? How about protecting consumers from other consumers (which is one of the things that Comcasts's throttling of BitTorrent did)?

I think Julian Sanchez had this about right: regulation may be required, but you need to regulate very narrowly, to address very narrow problems. The problem with Genachowski's speech is that it was very broad. That may be an artifact of it being a speech, but it certainly leaves open the door for broad regulation.

Honestly, I just don't follow this one.

It reads like a Randian dogwhistle to me. Isn't it a common belief among libertarians that a conscious desire to stifle innovation for its own sake exists in government?

@Rad -- Again, this is where the difference b/w open and neutral becomes key. No one wants to deny ISPs the ability to block harmful things. And the FCC isn't outlawing network management by any stretch.

Instead, they are channeling it in a more open, transparent, and constructive direction. Remember that Comcast blocked this stuff in secret and then lied about it -- that generates all sorts of downstream costs and problems.

Also, I'm not entirely sure I agree with your layer description -- I think of them as physical, Internet (IP protocol), transport (TCP) and application. comcasat is using control over physical layer to mess w/ IP layer -- it's not regulating application layer stuff

Also, what angered me about this post was the tone. I'll probably write about this later but it was a borderline anti-government screed. Openness debates haven't become Glenn Beck-politicized yet, but pieces like this make that more likely (note that I first even heard of this article b/c of a Glenn Reynolds approving link)

Also, I'm not entirely sure I agree with your layer description -- I think of them as physical, Internet (IP protocol), transport (TCP) and application. comcasat is using control over physical layer to mess w/ IP layer -- it's not regulating application layer stuff

If you want to use the internet protocol suite layering, it goes 1=datalink (equivalent to OSI layers 1 and 2), 2=internet (OSI layer 3), 3=transport (OSI layer 4), and 4=application (OSI layers 5, 6, and 7).

Now, what Comcast did to BitTorrent is to snoop--at the application layer (4), for BT upload requests and, upon seeing them, they jammed a TCP connection reset packet (at transport layer 3) back down to the node attempting to do the upload. This causes the TCP stack to kill the connection. I found a more complete explanation here.

At no point did they do anything at the physical layer, nor did they even do anything to the IP packets.

I apologize for the geekery, but it will hopefully give you a flavor for how complicated this is going to be to regulate.

I don't think merely describing the technical process makes it complicated at all. The upshot here is they blocked it. And their blocking was enabled by their physical control of infrastructure. You can't do that. And if there was disclosure and transparency, people like yourself would figure this out in a nanosecond and it'd be on every techblog.

How about protecting consumers from other consumers (which is one of the things that Comcasts's throttling of BitTorrent did)?

That seems like an odd description- Comcast has chosen to sell X bandwidth to comsumers, but have oversold their capacity. Thus, they attack one of the more heavily-used protocols to cut down on traffic.
This is like an airline overselling seats on a plane, then bumping passengers if too many show up. Not many people would buy the misrepresentation that this is being done to 'protect passengers.'

nb Im not against the concept of overselling, just the ludicrous concept that Im somehow attacking or defrauding Comcast or other Comcast customers by trying to use all of the bandwidth they've sold me.

Interesting that the only ISP mentioned here is Comcast.

Isn't their picture in the dictionary next to "Hideous Customer Service"?

They exhibit all the worst characteristics of a monopoly provider.

Why should the IPSs be given veto power over new applications and/or protocols on the Internet, just to protect them from the consequences of their own deceptive advertising and under-deployed networks (often built while either a monopoly, or with copious subsidies)?

Carleton, I'd buy your airline analogy if you had to pay by the bit, like you do on some mobile 3G plans. But the MSOs (cable operators) in particular have sold you a service, not a big bag o' capacity. That service has to be deployed statistically to be commercially viable. (NB: "commercially viable" = "deployable at a price where anybody would buy it", not "immensely profitable in a greedy, corporations-are-inherently-evil sort of way.")

Now, I'm going to tell you why BitTorrent is problematic, but before you all jump on me, I am not defending what Comcast did, i.e. nuking connections and not telling anybody about it. That said:

The web is set up on a very particular, highly asymmetric model. It assumes that small consumers mostly download data from considerably bigger producers of data (i.e. the servers of content providers). This has a huge impact on how traffic is engineered in the internet. It is the principal reason why your upload bitrate is considerably lower than your download bitrate.

BitTorrent turns that on its head. It turns small home systems into fairly hi-bitrate servers, which in turn gobble up a huge amount of the upload bandwidth of an MSO. To further complicate things, the DOCSIS networking that MSOs deploy at the neighborhood level is truly shared bandwidth, like ethernet. That means that your neighbor's BitTorrent node eats into your ability physically to put packets onto the cable. In short, BitTorrent may be a fine internet application, but as a web application it's a bad citizen.

Other non-web-like applications are real time VoIP and live video, as I ranted about on one of the previous threads. However, unlike BitTorrent, they simply stop working at the point that they become bad citizens, because they're not TCP-based. To get them to work properly, somebody has to pay for the QoS that's needed to run them. BitTorrent does use TCP, which means that it's perfectly happy to drive the network right up to the point of maximum utilization without ever quite causing congestive collapse.

Now, I'm not suggesting that BitTorrent be outlawed, and I'm certainly not suggesting that an MSO has the right to violate its SLA and deny service to an application without explicitly modifying that SLA and classifying P2P apps as something akin to some kind of malware. I am suggesting that an ISP is responsible for managing its network service so that it delivers the best value for the bulk of its customers. That simply is not possible while maintaining application neutrality. This idea that the internet is some kind of pristine medium that can act as a universal platform for any type of communications app that some 18-year-old can dream up in his basement is a fantasy. I'm just waiting to see whether the FCC shares the fantasy or not.

publius, I want to make sure I understand your position: Which of the following are you in favor of:


  1. That the FCC mandate transparency, where any traffic management policies are made public and auditable?

  2. That the FCC mandate neutrality so that there is no discrimination against any provider or consumer that produces or consumes the same type of data?

  3. That the FCC mandate application neutrality, where no type of data may be discriminated against?

Just to be clear, I'm in favor of #1 and #2, but not #3.

Extra credit: Which of the above do you think the FCC is in favor of? (I genuinely have no clue from the speech.)

Another way of lookig at BitTorrent is that makes transferring large files dramatically more efficient.

And the whole maximum utilization is a bit of a canard. The network has since its inception deal with congested traffic by slowing everything down together a bit. More broadly, there are widely-accepted standards on how to deal with it, even if I don't have the enginnering exactly right.

I don't think uploading is nearly a problem on the scale you're describing.

I'd agree, but I'm in favor of #3 too b/c I think (to extent a problem even exists, which I think is debatable), adding capacity cures all woes. Various engineers in the comcast proceeding explained why the whole p2p takes up maximum capacity isn't right.

But I think all of these are slightly more complex, esp. 1. No one is going to make ISPs turn over things that enable spammers, malicious hackers, etc. But there are ways to disclose what management practices you'll do, what congestion responses you'll take, etc.

I mean, I think the speech goes out of its way to signal that "hey, we're going to be flexible here,b ut you can't pull this egregious stuff."

I'm obviously veyr much enjoying the conversation, but I have to run. Will be back on later

BitTorrent doesn't make transferring large files more efficient; it makes distributing those files to lots of different nodes more efficient. (It makes distributing those files vastly more efficient if uploading them to a central server would violate that server's EULA...)

There are widely-accepted standards on how TCP deals with congestion, which is why BitTorrent actually works. If you go outside TCP, you're toast.

Uploading is only a problem if you've engineered your network for web usage. I'm on Time Warner, and my download speed is 15 Mbps and my upload speed is 1 Mbps. BitTorrent is roughly symmetric in its download/upload usage. How well do you think that will work?

"adding capacity cures all woes"

The heart of the problem, too many people just believe this, and that adding capacity is free(or, if not, should be required anyway).

I'm on Time Warner, and my download speed is 15 Mbps and my upload speed is 1 Mbps. BitTorrent is roughly symmetric in its download/upload usage. How well do you think that will work?

Do you actually get 15/1? As in, can you download at 15 Mbps for a half hour? Because when I was last with Comcast, I could download at 6 Mbps for about 30 seconds, after which throughput would fall by a factor of 10 or 20.

Marty: "The heart of the problem, too many people just believe this, and that adding capacity is free (or, if not, should be required anyway)."

A) People "believe this" (that adding network capacity works just as well as QoS measures) because it is true.

B) Nobody thinks it is "free". They think that in a capitalist society the regulated monopolies/duopolies that they are forced to deal with for Internet access ought to be spending their money on network upgrades so as to expand service availability, and if they refuse to do so, they ought to lose their privileged position as monopoly providers to someone who will. Comcast and AT&T do not operate in truly competitive markets; see Econ 101 on "natural monopolies" if you've forgotten how that works.

When we last saw Marty, he was claiming that Comcast needed to charge internet companies rent to allow Comcast customers to access their servers because people wouldn't tolerate Comcast raising their fees, so they should be allowed to charge others who were not service subscribers because... well, just because.

If comcast feels that its business model is not sustainable under the constraints of what providing internet service entails, they are free to develop their own, closed network and start charging customers and application developers access to use it. This isn't a difficult concept.

When we last saw Marty he was claiming that the assumption that Comcast should be treated the same as the top tier layer of the internet providers was incorrect in specific cases, VOIP and (or P2P in general).

He quit posting on this because RadMod was doing such a better job of actually explaining why.

So he will now just point out that RadMod has made the point much better the he could.

Thanks.

RM: You think the companies aren't offering you a rate? They quote rates in all of their ads, plans, etc. That they overpromise on what their network can maintain, because they designed the plans before widespread video, BitTorrent, etc became available doesn't justify giving the ISPs veto power over new applications on the net, and veto power over competitors products used over the network.

Nor does it really justify their claiming X mb up and Y mb down, then putting in tiny print that the quality may vary depending on "network conditions" where the "conditions" that may degrade your performance is from anybody else in your neighborhood using the net at all.

Seriously, when they had the opportunity before, the local phone monopolies strangled independent DSL operations, and the telecom companies have all done their best to kill municipal wifi. These are not hypothetical violations, this is a pattern of bad acts by these very companies.

Just in case any one is interested, I posted a fairly detailed explanation of some of the technical considerations that go into traffic engineering, and why application neutrality is going to be problematic, over on my blog. It's quite long--way too long to be a comment here--and I have no idea if I've done a decent job explaining this for motivated non-geeks. It's there if you want it.

Turb, no, I don't get anything like 15 down and 1 up, but I'm pretty sure that they've slotted DOCSIS with that approximate ratio, which is the point I was trying to make. It will never get better than the two advertised numbers.

Nate, I'm having a middle-aged moment and can't remember the term for the program that gave open access to third-party local exchange carriers, but those third-party LECs might have survived if the Bush Administration had enforced the existing regulations.

On the other hand, it's an incredibly stupid business model: You're a startup LEC, and you're not only entering a market that by definition has already been commoditized, but you're paying rent on somebody else's physical plant, which is the only non-commodity component of the business. And these guys expected to pay back their venture capitalists?

As for the Wifi stuff, I predict it will be back with vengeance, to the cheers of the mobile operators. They've stepped in some nasty stuff with 3G data swamping their voice networks and their customers are screaming mad. They're dying to offload traffic.

Look, I'm not saying that the ISPs aren't going to cut your heart out and sell it back to you if you let them get away with it. I'm merely saying that you can't legislate pi equal to 3.0 and expect very many round things to survive. There are some real technical constraints that could become even worse with bad regulation. I want effective, but minimal regulation. Other than app neutrality, I'm a happy camper.

RadicalModerate, before I spend time reading your lengthy technical treatise, do you have any explanation for why you wrote such absurd things about the terrible threat that wideband VOIP poses to the internet? I mean, do you actually think that the human auditory system would benefit from significantly more bandwidth (i.e., much more than a factor of 2-4) than the 8 kHz sampling that the phone system employs?

do you have any explanation for why you wrote such absurd things about the terrible threat that wideband VOIP poses to the internet? I mean, do you actually think that the human auditory system would benefit from significantly more bandwidth (i.e., much more than a factor of 2-4) than the 8 kHz sampling that the phone system employs?

I don't think it's a terrible threat, but I'll tell you what I do know. I listen to people all day long on G.711 VoIP inside an enterprise, on VPNs that feed into one of the best traffic-engineered networks in the world--and it breaks up at the edge.

Now, G.711 isn't wideband, but it is 64 Kbps, which is about what wideband is when it's deployed. Therefore, I expect that there will be problems in the public net.

Now, as for the human auditory system, I'll only state that you've been using telephones for your whole life and you're used to it sounding like a telephone. True wideband (not just wideband encoding, but wideband mics and speakers) sounds... different. Better. A lot better.

Having said all that, I expect much worse problems with video than I do with wideband audio.

I don't think it's a terrible threat, but I'll tell you what I do know.

Really? Then why did you write "VoIP is not as much of a problem (although wideband VoIP likely will be)"?

I listen to people all day long on G.711 VoIP inside an enterprise, on VPNs that feed into one of the best traffic-engineered networks in the world--and it breaks up at the edge.

Ah yes, this must be one of the best traffic engineered networks in the world. I mean, sure, it is apparently nowhere near as good as a crappy Skype connection running through residential DSL, but it is one of the best traffic-engineered networks in the world. Really it is!

Now, as for the human auditory system, I'll only state that you've been using telephones for your whole life and you're used to it sounding like a telephone. True wideband (not just wideband encoding, but wideband mics and speakers) sounds... different. Better. A lot better.

A lot better? Really? Look, I'm sure you have some very fancy microphones. And super fancy speakers. But some of us (like me) actually went to school and studied things like human auditory response and signal processing and the Nyquist sampling theorem. I know, that boring stuff can't possibly compare with your amazing microphones and your incredibly well engineered network, but, you know, wideband speech codecs sample at 16 kHz. That's twice as fast telephone class codecs. A factor of two over an already tiny bandwidth use will not hurt the internet. If that's not obvious to you at a deep fundamental level, you don't have any technical insights to offer to anyone.

ATandT sent a letter to the FCC protesting Google's blocking of telephone calls from consumers that use its Google Voice service to call phone numbers with inflated access charges in certain rural areas. By blocking these calls, Google reduces its access expenses, giving it an advantage phone carriers are prevented from enjoying and thus skewering the competition principles in U.S. network neutrality laws, ATandT claims. Google argued that Google Voice must not be accorded the same treatment as services from phone carriers.
[...]

ATandT Tells FCC Google Voice Violates Network Neutrality Laws

By blocking these calls, Google reduces its access expenses, giving it an advantage phone carriers are prevented from enjoying and thus skewering the competition principles in U.S. network neutrality laws, ATandT claims.

This one will be interesting. It's almost certain that Google is blocking this stuff at their PSTN egress gateways, by configuring their dialplan to omit the numbers they don't want to pay for. You could argue (somewhat disingenuously) that the IP access is completely neutral, and it's only when the signaling hits the PSTN dialplan that the call gets blocked.

I assume that everybody knows that the basis of Google's complaint against Apple and AT&T was that the AppStore won't accept any application that tries to do VoIP over 3G. Three main reasons for this:

  1. AT&T is terrified that VoIP/3G will completely destroy their network (that is, destroy it more that 3G data already has destroyed it).

  2. AT&T is terrified that allowing VoIP will commoditize their network, reducing them to a lowly ISP (which it would).

  3. AT&T and Apple are scared that VoIP apps over 3G will create a flood of applications that programmatically make and receive calls, which sucks all the value out of the iPhone UI, which they're hoping (probably vainly) will give them an edge over other smartphone vendors.
This is all almost as much fun as last January's behind-the-scenes lobbying fight where WiMax and 4G beat each other over the head with the delay in transitioning to digital TV broadcast.

"It bothers me (somewhat) that 'libertarian' has come to mean 'social-Darwinism on the make'."

It bothers me too. Proudhon, one of the founders of "libertarian" thinking, was a socialist anarchist and would have been disgusted by these people.

The comments to this entry are closed.