Is Comcast discriminating against BitTorrent?

by on October 20, 2007 · 47 comments

The AP reports today the results of an investigation it conducted on Comcast’s “traffic shaping” practices as they relate to BitTorrent. The bottom line, if the AP is correct, is that Comcast interferes with packets coming from both ends of a BitTorrent communication. Comcast allegedly inserts messages pretending to be one or the other end requesting that the transmission be reset. Susan Crawford has a technical explanation on her blog.

If this is a consistent policy, this is much worse than the meaningless one-off snafus such as Madison River, Pearl Jam, or NARAL. While this is technically legal, and should always be, it’s a bit indefensible. No doubt Comcast and every other access provider should have the ability to manage their networks to ensure that a minority of users doesn’t slow down or increase costs for the majority. However, they should be transparent about what they do.

As the AP reports it (and I’m really looking forward to clarification), “Comcast’s technology kicks in, though not consistently, when one BitTorrent user attempts to share a complete file with another user.” If that means any BitTorrent user, even if they’re not a heavy user, then the policy seems over-broad to me. In its acceptable use policy,1 Comcast reserves the right to take any measures it deems necessary to deal with subscribers who use too much bandwidth (although how much is too much is not clearly defined). But if the AP is right, this is targeting a specific application, not specific users.

What this all points out to me, however, is that we don’t need regulation prohibiting these kinds of network management practices. The problem is not the practice, but the lack of disclosure, and as Google’s Andrew McLaughlin has said, it’s more of an FTC issue than an FCC one. The other issue this brings up is Adam’s favorite: Why not just have a Ramsey two-part tariff style metering after instead of interfering with legitimate applications?


  1. See the relevant portions of the acceptable use policy here.

  • http://bennett.com/blog Richard Bennett

    There’s tremendous confusion about this issue, much of it intentional, and the rest of it a lack of understanding of network operating principles on the network critics such as Crawford.

    Let’s posit that there’s a reasonable form of network management, which operates like this:

    1) When demand for network bandwidth on shared facilities is low, every user gets as much as he wants.

    2) When demand for network bandwidth exceeds supply, every user is allocated bandwidth equitably.

    3) “Equitable” allocation means something like this: every user requesting less than the average per-user available bandwidth gets what he requests, and those who request more get additional bandwidth when it’s available.

    That’s a reasonable algorithm implemented in a number of commercial systems today, and please note that’s it’s content- and viewpoint-neutral.

    And also note that as a practical matter it’s only necessary to examine BitTorrent traffic on the typical ISP network to implement it, because (as a practical matter) all the excess demand for bandwidth comes from BitTorrent.

    And also note that the slickest way to throttle BitTorrent is to limit the number of uploads a given user can offer, which is exactly what the TCP Reset (RST flag) spoofing does.

    Given all of that, is there anything to see here other than an ISP applying reasonable principles by reasonable means?

    I conducted my own experiment on Comcast and found that I’m not impeded from uploading or downloading either legal or illegal files. Go see a screen shot.

    Moral of the story: don’t believe everything Susan Crawford and her ilk tell you about the Internet.

  • http://bennett.com/blog Richard Bennett

    There’s tremendous confusion about this issue, much of it intentional, and the rest of it a lack of understanding of network operating principles on the network critics such as Crawford.

    Let’s posit that there’s a reasonable form of network management, which operates like this:

    1) When demand for network bandwidth on shared facilities is low, every user gets as much as he wants.

    2) When demand for network bandwidth exceeds supply, every user is allocated bandwidth equitably.

    3) “Equitable” allocation means something like this: every user requesting less than the average per-user available bandwidth gets what he requests, and those who request more get additional bandwidth when it’s available.

    That’s a reasonable algorithm implemented in a number of commercial systems today, and please note that’s it’s content- and viewpoint-neutral.

    And also note that as a practical matter it’s only necessary to examine BitTorrent traffic on the typical ISP network to implement it, because (as a practical matter) all the excess demand for bandwidth comes from BitTorrent.

    And also note that the slickest way to throttle BitTorrent is to limit the number of uploads a given user can offer, which is exactly what the TCP Reset (RST flag) spoofing does.

    Given all of that, is there anything to see here other than an ISP applying reasonable principles by reasonable means?

    I conducted my own experiment on Comcast and found that I’m not impeded from uploading or downloading either legal or illegal files. Go see a screen shot.

    Moral of the story: don’t believe everything Susan Crawford and her ilk tell you about the Internet.

  • http://www.jerrybrito.com Jerry Brito

    Richard, Thanks for that explanation. It’s really useful, and in my post I tried to convey that I really want to know all the facts. That said, maybe you can answer a couple other questions for me.

    Your own tests aside, am I wrong in understanding the press accounts as meaning that Comcast is targetting BitTorrent generally and not just the BitTorrent use of heavy users? I understand that BitTorrent is, in practical terms responsible for most of the abuse, but why should that affect the BitTorrent use of non-abusive customers?

    Second, I’m not disputing that this may well be an efficient network management practice, my concern is that it’s not disclosed. Shouldn’t bandwidth caps be more clearly defined? Shouldn’t consumers know how some of their applications are being affected?

  • http://www.manifestdensity.net Tom

    Richard, you seem to misunderstand what Comcast’s doing. In most areas the RSTs start firing after the user begins seeding. Your example doesn’t show any completed downloads.

    And this makes no sense at all:

    And also note that the slickest way to throttle BitTorrent is to limit the number of uploads a given user can offer, which is exactly what the TCP Reset (RST flag) spoofing does.

    It’s not clear that this is an efficient way to limit Bittorrent use. It IS an extremely sneaky and hard-to-detect means of throttling traffic, however. This is consistent with Comcast’s other traffic-limiting behavior, like enforcing bandwidth caps that they refuse to disclose. The idea seems to be to scare away unprofitable consumers without actually admitting that their service is limited.

    As Jerry points out, this is a pretty lousy way to treat your customers. Comcast should disclose what they think their customers’ subscription fees do and don’t entitle them to. Unfortunately even if they do, most users will only have one alternative for broadband. Is that good enough? I hope so.

  • http://bennett.com/blog Richard Bennett

    Jerry, the press reports that Comcast is targeting BitTorrent turned out to be wrong, as the EFF is now reporting that Gnutella and Notes are also capped. It appears that Comcast is managing P2P uploads generally.

    We don’t know anything about the algorithm behind the throttling of P2P. It could be that it’s only active when the network is heavily loaded, or maybe it’s only active for users that have exceeded a quota, or maybe that it’s only active against uploads that leave the Comcast network; we do know that it’s not simply a ban on P2P in all places at all times. And until we do know the scope of what Comcast is doing (something we can’t gather from a couple of test points on a network of millions of customers) we don’t know if it’s worth getting excited about.

    Disclosure is an interesting question, because most of Comcast’s customers aren’t able to understand a disclosure concerning TCP Resets and P2P uploads, I dare say. At the extreme of simplicity, I suppose they could disclose something like this: “The laws of physics do not permit Comcast to deploy infinite bandwidth. We regret the limitations in the speed of light and will endeavor to have them repealed by legislative action in the Congress of God’s Laws. Thanks for playing,” but some folks would be insulted.

    I take it that they do disclose a vague general policy against bandwidth hogs, and that’s enough to satisfy me.

    Packet networks like the Internet are built on the assumption of good user behavior, and when this doesn’t pan out they provide few administrative tools for enforcement of norms. One method is simply to drop TCP segments, but that simply causes them to be retransmitted. Spoofing Reset is actually the only means of cutting off a flow at its root, and is therefore the most efficient means of throttling P2P.

    Crawford implies this is something like identity theft, and she should be spanked for that because it’s a totally bogus analogy. Comcast is trying to ensure that I can access my e-mail while my neighbor’s teenage son is sharing the Paris Hilton sex tape, and that’s fine by me. In fact, our Democracy depends on their doing this.

  • http://jerrybrito.com jerrybrito

    Richard, Thanks for that explanation. It’s really useful, and in my post I tried to convey that I really want to know all the facts. That said, maybe you can answer a couple other questions for me.

    Your own tests aside, am I wrong in understanding the press accounts as meaning that Comcast is targetting BitTorrent generally and not just the BitTorrent use of heavy users? I understand that BitTorrent is, in practical terms responsible for most of the abuse, but why should that affect the BitTorrent use of non-abusive customers?

    Second, I’m not disputing that this may well be an efficient network management practice, my concern is that it’s not disclosed. Shouldn’t bandwidth caps be more clearly defined? Shouldn’t consumers know how some of their applications are being affected?

  • http://www.manifestdensity.net Tom

    Richard, you seem to misunderstand what Comcast’s doing. In most areas the RSTs start firing after the user begins seeding. Your example doesn’t show any completed downloads.

    And this makes no sense at all:

    And also note that the slickest way to throttle BitTorrent is to limit the number of uploads a given user can offer, which is exactly what the TCP Reset (RST flag) spoofing does.

    It’s not clear that this is an efficient way to limit Bittorrent use. It IS an extremely sneaky and hard-to-detect means of throttling traffic, however. This is consistent with Comcast’s other traffic-limiting behavior, like enforcing bandwidth caps that they refuse to disclose. The idea seems to be to scare away unprofitable consumers without actually admitting that their service is limited.

    As Jerry points out, this is a pretty lousy way to treat your customers. Comcast should disclose what they think their customers’ subscription fees do and don’t entitle them to. Unfortunately even if they do, most users will only have one alternative for broadband. Is that good enough? I hope so.

  • http://bennett.com/blog Richard Bennett

    Jerry, the press reports that Comcast is targeting BitTorrent turned out to be wrong, as the EFF is now reporting that Gnutella and Notes are also capped. It appears that Comcast is managing P2P uploads generally.

    We don’t know anything about the algorithm behind the throttling of P2P. It could be that it’s only active when the network is heavily loaded, or maybe it’s only active for users that have exceeded a quota, or maybe that it’s only active against uploads that leave the Comcast network; we do know that it’s not simply a ban on P2P in all places at all times. And until we do know the scope of what Comcast is doing (something we can’t gather from a couple of test points on a network of millions of customers) we don’t know if it’s worth getting excited about.

    Disclosure is an interesting question, because most of Comcast’s customers aren’t able to understand a disclosure concerning TCP Resets and P2P uploads, I dare say. At the extreme of simplicity, I suppose they could disclose something like this: “The laws of physics do not permit Comcast to deploy infinite bandwidth. We regret the limitations in the speed of light and will endeavor to have them repealed by legislative action in the Congress of God’s Laws. Thanks for playing,” but some folks would be insulted.

    I take it that they do disclose a vague general policy against bandwidth hogs, and that’s enough to satisfy me.

    Packet networks like the Internet are built on the assumption of good user behavior, and when this doesn’t pan out they provide few administrative tools for enforcement of norms. One method is simply to drop TCP segments, but that simply causes them to be retransmitted. Spoofing Reset is actually the only means of cutting off a flow at its root, and is therefore the most efficient means of throttling P2P.

    Crawford implies this is something like identity theft, and she should be spanked for that because it’s a totally bogus analogy. Comcast is trying to ensure that I can access my e-mail while my neighbor’s teenage son is sharing the Paris Hilton sex tape, and that’s fine by me. In fact, our Democracy depends on their doing this.

  • http://www.manifestdensity.net Tom

    I think you’re incorrect, Richard. Look, it’s great that the EFF is doing work to determine if other protocols are affected, but they’ve also confirmed that Comcast’s throttling responds to BT-specific network events. I wouldn’t be at all surprised if they’re capping other forms of P2P, too — the Sandvine product that they’re reportedly using targets other forms of P2P besides BT (it’s odd that it’s affecting Lotus Notes, which is not P2P, but I’m not at all surprised that some applications are being unintentionally broken by this system).

    And until we do know the scope of what Comcast is doing (something we can’t gather from a couple of test points on a network of millions of customers) we don’t know if it’s worth getting excited about.

    You seem to be under the impression that the AP broke this story. They didn’t — it’s been discussed in the BroadbandReports forums for months. Torrentfreak’s been following it closely, and Wired News picked it up in August. Many, many people have confirmed that they’re affected. Comcast is applying the measures erratically, but they seem to be widespread. I’m on Verizon, but I’ve got a neighbor who’s now unable to seed; check out BBReports and you’ll find hundreds more.

    Packet networks like the Internet are built on the assumption of good user behavior, and when this doesn’t pan out they provide few administrative tools for enforcement of norms. One method is simply to drop TCP segments, but that simply causes them to be retransmitted. Spoofing Reset is actually the only means of cutting off a flow at its root, and is therefore the most efficient means of throttling P2P.

    A few things. First, relatively high network utilization does not constitute a deviation from “good user behavior” — the user has no way of knowing how loaded his or her network segment is, and can’t reasonably be expected to manage Comcast’s bandwidth for them. The second is flatly incorrect — many if not most ISPs already use per-user throttling to provide high burst speeds that are useful for web browsing, yet manage to put on the brakes for sustained traffic. There are many tools available for handling these sorts of issues.

    This is not super-high-tech: I’ve got throttling turned on for unauthenticated users here on my apartment’s WLAN; at work we’ve got per-protocol throttling to provide priority for VoIP traffic. In both cases the network management is done without breaking TCP connections (and done with cheap hardware and free software, I might add — Comcast can’t plead poverty when they’re buying more expensive proprietary products like Sandvine). Forging reset packets is not the sort of thing that network administrators typically do to shape traffic. It’s a very, very blunt instrument.

    I’m not trying to imply that Comcast should allow unrestricted BT traffic. But they should disclose what they’re doing, and use less destructive means of managing their network.

  • http://bennett.com/blog Richard Bennett

    The essence of Tom’s argument seems to be that TCP Resets are bad, and dropping TCP segments is good (that being the only other known way of “throttling” TCP traffic.)

    When we get to this point in the argument, everyone who isn’t a network engineer has to bow out. The fact that somebody believes one form of management is more “typical” than the other says nothing about its virtue, or lack thereof.

    Seriously.

  • Lewis Baumstark

    Comcast (likely) has every legal right to throttle, reshape packets, etc. as it sees fit, without notification or disclosure.

    However, they richly deserve the court-of-public-opinion thrashing they get when they do that along with any lost business that results from it.

  • http://www.manifestdensity.net Tom

    I think you’re incorrect, Richard. Look, it’s great that the EFF is doing work to determine if other protocols are affected, but they’ve also confirmed that Comcast’s throttling responds to BT-specific network events. I wouldn’t be at all surprised if they’re capping other forms of P2P, too — the Sandvine product that they’re reportedly using targets other forms of P2P besides BT (it’s odd that it’s affecting Lotus Notes, which is not P2P, but I’m not at all surprised that some applications are being unintentionally broken by this system).

    And until we do know the scope of what Comcast is doing (something we can’t gather from a couple of test points on a network of millions of customers) we don’t know if it’s worth getting excited about.

    You seem to be under the impression that the AP broke this story. They didn’t — it’s been discussed in the BroadbandReports forums for months. Torrentfreak’s been following it closely, and Wired News picked it up in August. Many, many people have confirmed that they’re affected. Comcast is applying the measures erratically, but they seem to be widespread. I’m on Verizon, but I’ve got a neighbor who’s now unable to seed; check out BBReports and you’ll find hundreds more.

    Packet networks like the Internet are built on the assumption of good user behavior, and when this doesn’t pan out they provide few administrative tools for enforcement of norms. One method is simply to drop TCP segments, but that simply causes them to be retransmitted. Spoofing Reset is actually the only means of cutting off a flow at its root, and is therefore the most efficient means of throttling P2P.

    A few things. First, relatively high network utilization does not constitute a deviation from “good user behavior” — the user has no way of knowing how loaded his or her network segment is, and can’t reasonably be expected to manage Comcast’s bandwidth for them. The second is flatly incorrect — many if not most ISPs already use per-user throttling to provide high burst speeds that are useful for web browsing, yet manage to put on the brakes for sustained traffic. There are many tools available for handling these sorts of issues.

    This is not super-high-tech: I’ve got throttling turned on for unauthenticated users here on my apartment’s WLAN; at work we’ve got per-protocol throttling to provide priority for VoIP traffic. In both cases the network management is done without breaking TCP connections (and done with cheap hardware and free software, I might add — Comcast can’t plead poverty when they’re buying more expensive proprietary products like Sandvine). Forging reset packets is not the sort of thing that network administrators typically do to shape traffic. It’s a very, very blunt instrument.

    I’m not trying to imply that Comcast should allow unrestricted BT traffic. But they should disclose what they’re doing, and use less destructive means of managing their network.

  • http://bennett.com/blog Richard Bennett

    The essence of Tom’s argument seems to be that TCP Resets are bad, and dropping TCP segments is good (that being the only other known way of “throttling” TCP traffic.)

    When we get to this point in the argument, everyone who isn’t a network engineer has to bow out. The fact that somebody believes one form of management is more “typical” than the other says nothing about its virtue, or lack thereof.

    Seriously.

  • Lewis Baumstark

    Comcast (likely) has every legal right to throttle, reshape packets, etc. as it sees fit, without notification or disclosure.

    However, they richly deserve the court-of-public-opinion thrashing they get when they do that along with any lost business that results from it.

  • http://bennett.com/blog Richard Bennett

    Given that Comcast’s actions are reasonable as well as beneficial to the vast majority of their customers, what sort of thrashing do you propose the court of public opinion ought to deliver, Lewis, one that says: “I don’t understand what you’re doing, waaaahhhh!!!!.”

  • http://bennett.com/blog Richard Bennett

    Given that Comcast’s actions are reasonable as well as beneficial to the vast majority of their customers, what sort of thrashing do you propose the court of public opinion ought to deliver, Lewis, one that says: “I don’t understand what you’re doing, waaaahhhh!!!!.”

  • http://zgp.org/~dmarti/ Don Marti

    A simple workaround to the forged RST has been around for a while.

  • http://zgp.org/~dmarti/ Don Marti

    A simple workaround to the forged RST has been around for a while.

  • Lewis Baumstark

    “I don’t understand what you’re doing, waaaahhhh!!!!.”

    Richard, that’s very close to what’s going on. Customers clearly aren’t able to do what they think they should. Since Comcast’s policy is vague and since, as you imply, not everyone is a network engineer, what would you expect public reaction to be? Roll over and take it? No, clearly people will speculate, discuss, and beat up on Comcast until Comcast comes clean about what they are doing.

  • Lewis Baumstark

    “I don’t understand what you’re doing, waaaahhhh!!!!.”

    Richard, that’s very close to what’s going on. Customers clearly aren’t able to do what they think they should. Since Comcast’s policy is vague and since, as you imply, not everyone is a network engineer, what would you expect public reaction to be? Roll over and take it? No, clearly people will speculate, discuss, and beat up on Comcast until Comcast comes clean about what they are doing.

  • http://bennett.com/blog Richard Bennett

    My own Comcast experiment in seeding has been running for 24 hours now, during which time I have been able to successfully seed an illegal file after its downloading was complete. I seeded it to customers of AT&T, Verizon, Cable 1, Wide Open West, APNIC (Asia-Pacific), and RIPE (Amsterdam.)

    Based on my experience, the reports that Comcast bans or blocks BitTorrent are grossly exaggerated. They do appear to be *rationing* upload bandwidth for P2P use, which is a totally sensible thing to do given the underlying mechanisms in cable modems. In fact, I would go so far as to say that if they didn’t do something like this, they would be irresponsible.

    The vast majority of Comcast customers are not bandwidth hogs and BitTorrent seeds, after all.

    So I give Comcast a pat on the back and Gold Star.

  • http://bennett.com/blog Richard Bennett

    My own Comcast experiment in seeding has been running for 24 hours now, during which time I have been able to successfully seed an illegal file after its downloading was complete. I seeded it to customers of AT&T;, Verizon, Cable 1, Wide Open West, APNIC (Asia-Pacific), and RIPE (Amsterdam.)

    Based on my experience, the reports that Comcast bans or blocks BitTorrent are grossly exaggerated. They do appear to be *rationing* upload bandwidth for P2P use, which is a totally sensible thing to do given the underlying mechanisms in cable modems. In fact, I would go so far as to say that if they didn’t do something like this, they would be irresponsible.

    The vast majority of Comcast customers are not bandwidth hogs and BitTorrent seeds, after all.

    So I give Comcast a pat on the back and Gold Star.

  • http://www.manifestdensity.net Tom

    Richard, why are you so convinced that throttling has to involve the destruction of data (and why are you using scare quotes around the word throttling?)? Have a look at the Wikipedia page on the subject. As it makes clear, throttling can be achieved by queuing requests, allowing them through at an appropriate rate. There’s no need to throw away the packet or tell the client that the connection has broken. It’s destructive and unnecessary. I don’t know why you keep insisting that this is the only way to implement traffic shaping.

  • http://www.manifestdensity.net Tom

    Richard, why are you so convinced that throttling has to involve the destruction of data (and why are you using scare quotes around the word throttling?)? Have a look at the Wikipedia page on the subject. As it makes clear, throttling can be achieved by queuing requests, allowing them through at an appropriate rate. There’s no need to throw away the packet or tell the client that the connection has broken. It’s destructive and unnecessary. I don’t know why you keep insisting that this is the only way to implement traffic shaping.

  • http://bennett.com/blog Richard Bennett

    Thanks for pointing me to Wikipedia, Tom, it’s always good for a laugh. I’ve been known to design and implement traffic management software from time to time myself, and I try to stay current.

    As you get your information on the subject from Wikipedia, it’s not surprising that you’re confused about the method Comcast (and 100 other ISPs around the world) use. Cable modems, you see, have very limited upload bandwidth, due to the multiple transmitter problem. They employ a slotting system to arbitrate accesses. So the weak link in the entire Comcast network is the first hop. The queuing technique you propose can only be employed above the first hop, so it doesn’t help. Furthermore, even if they could delay the first hop, BT responds to delay my opening more TCP connections, so ultimately it still wouldn’t help.

    But closing connections causes BT to look elsewhere for free bandwidth, and that’s just what Comcast (and the vast majority of its customers, doncha know) want it to do.

    From the standpoint of a cable network provider, P2P is indistinguishable from a spambot: it takes over users’ computers and hijacks network bandwidth for its own purposes. It needs to be throttled, by any means necessary.

    Gold Star to Comcast.

  • http://bennett.com/blog Richard Bennett

    Thanks for pointing me to Wikipedia, Tom, it’s always good for a laugh. I’ve been known to design and implement traffic management software from time to time myself, and I try to stay current.

    As you get your information on the subject from Wikipedia, it’s not surprising that you’re confused about the method Comcast (and 100 other ISPs around the world) use. Cable modems, you see, have very limited upload bandwidth, due to the multiple transmitter problem. They employ a slotting system to arbitrate accesses. So the weak link in the entire Comcast network is the first hop. The queuing technique you propose can only be employed above the first hop, so it doesn’t help. Furthermore, even if they could delay the first hop, BT responds to delay my opening more TCP connections, so ultimately it still wouldn’t help.

    But closing connections causes BT to look elsewhere for free bandwidth, and that’s just what Comcast (and the vast majority of its customers, doncha know) want it to do.

    From the standpoint of a cable network provider, P2P is indistinguishable from a spambot: it takes over users’ computers and hijacks network bandwidth for its own purposes. It needs to be throttled, by any means necessary.

    Gold Star to Comcast.

  • wirelessman

    Richard, you seem to be a bit deliberately obtuse here. You also suggest that this discussion is in the domain of network engineers and since I am one, I’ll wade in. I’m specifically involved in designing scheduling algorithms on shared last mile broadband links (uplink and downlink) for consumer access, so this is right up my alley. The specific transport technology I deal with is wireless (HSDPA and WiMAX) which is much more bandwidth constrained than cable and hence suffers more from bandwidth hogs like P2P users.

    The notion that throttling requires throwing away TCP packets is, as Tom says, incorrect. In any contention situation, you by definition have more user traffic than you can send over the link (up or down). If you didn’t, there wouldn’t be contention and you wouldn’t have a problem. In this situation, the scheduler gives each user less traffic than they’d like (i.e. it doesn’t empty their queues). In peak busy hour in broadband networks, nobody ever gets the advertised raw maximum link speed because the scheduler is dishing out a fraction of the pipe to each user.

    This behaviour is perfectly benign to TCP, which sees a lower bandwidth link with potentially increased latency. The TCP congestion control mechanisms ave no problem adapting to these variable link characteristics. This isn’t to say that the occasional packet doesn’t get dropped because TCP sends to many packets just as the link speed drops and queues overflow. These things happen and there are re-transmission mechanisms in TCP to deal with it.

    So all the carrier has to do is throttle BT users. Scheduling algorithms can even keep longer term stats on individual user’s usage so that the heavy users are throttled more than light users in contention situations (i.e. you end up being unfair to P2P users). These are straightforward mechanisms that are cheap, easy and most importantly, don’t interfere with traffic packets. What Comcast is doing is a sort of man-in-the-middle attack where they’re essentially spoofing traffic.

    You suggest that this is a perfectly reasonable approach for Comcast and throw a nice piece of snark at those suggesting disclosure: “The laws of physics do not permit Comcast to deploy infinite bandwidth. We regret the limitations in the speed of light and will endeavor to have them repealed by legislative action in the Congress of God’s Laws. Thanks for playing,”. There are less inane options. All Comcast has to do is impose bandwidth caps and actually disclose them. A cap of 100 GBytes per month would allow a BT user to download continuously at 40 kbps every second of every day…not an incredibly onerous load on the system. Or you could even do something as clever as have bandwidth caps in peak times and unlimited bandwidth in off-peak times (e.g. midnight – 8 AM). If users are aware of such policies, they can plan their use of P2P accordingly.

    So there are many reasonable approaches to solve the problem without resorting to interfering with the transport protocols themselves. The question is whether Comcast (and most other ISPs) can get past the “Unlimited usage” rhetoric of their marketing and be up front with customers about the reality of their network capabilities. Forget about P2P traffic, there’s a ton of other stuff out there that the networks can’t handle if everyone does it at once (i.e. VoIP, streaming audio and video, legal music downloads, video chat, etc). Maybe this isn’t about benefiting most Comcast consumers and is instead about satisfying third parties such as the RIAA and MPAA.

  • wirelessman

    Richard, you seem to be a bit deliberately obtuse here. You also suggest that this discussion is in the domain of network engineers and since I am one, I’ll wade in. I’m specifically involved in designing scheduling algorithms on shared last mile broadband links (uplink and downlink) for consumer access, so this is right up my alley. The specific transport technology I deal with is wireless (HSDPA and WiMAX) which is much more bandwidth constrained than cable and hence suffers more from bandwidth hogs like P2P users.

    The notion that throttling requires throwing away TCP packets is, as Tom says, incorrect. In any contention situation, you by definition have more user traffic than you can send over the link (up or down). If you didn’t, there wouldn’t be contention and you wouldn’t have a problem. In this situation, the scheduler gives each user less traffic than they’d like (i.e. it doesn’t empty their queues). In peak busy hour in broadband networks, nobody ever gets the advertised raw maximum link speed because the scheduler is dishing out a fraction of the pipe to each user.

    This behaviour is perfectly benign to TCP, which sees a lower bandwidth link with potentially increased latency. The TCP congestion control mechanisms ave no problem adapting to these variable link characteristics. This isn’t to say that the occasional packet doesn’t get dropped because TCP sends to many packets just as the link speed drops and queues overflow. These things happen and there are re-transmission mechanisms in TCP to deal with it.

    So all the carrier has to do is throttle BT users. Scheduling algorithms can even keep longer term stats on individual user’s usage so that the heavy users are throttled more than light users in contention situations (i.e. you end up being unfair to P2P users). These are straightforward mechanisms that are cheap, easy and most importantly, don’t interfere with traffic packets. What Comcast is doing is a sort of man-in-the-middle attack where they’re essentially spoofing traffic.

    You suggest that this is a perfectly reasonable approach for Comcast and throw a nice piece of snark at those suggesting disclosure: “The laws of physics do not permit Comcast to deploy infinite bandwidth. We regret the limitations in the speed of light and will endeavor to have them repealed by legislative action in the Congress of God’s Laws. Thanks for playing,”. There are less inane options. All Comcast has to do is impose bandwidth caps and actually disclose them. A cap of 100 GBytes per month would allow a BT user to download continuously at 40 kbps every second of every day…not an incredibly onerous load on the system. Or you could even do something as clever as have bandwidth caps in peak times and unlimited bandwidth in off-peak times (e.g. midnight – 8 AM). If users are aware of such policies, they can plan their use of P2P accordingly.

    So there are many reasonable approaches to solve the problem without resorting to interfering with the transport protocols themselves. The question is whether Comcast (and most other ISPs) can get past the “Unlimited usage” rhetoric of their marketing and be up front with customers about the reality of their network capabilities. Forget about P2P traffic, there’s a ton of other stuff out there that the networks can’t handle if everyone does it at once (i.e. VoIP, streaming audio and video, legal music downloads, video chat, etc). Maybe this isn’t about benefiting most Comcast consumers and is instead about satisfying third parties such as the RIAA and MPAA.

  • http://bennett.com/blog Richard Bennett

    Wirelessman, you’re assuming that the approach to traffic scheduling that you’re familiar with is available on DOCSIS networks. Polled networks, such as 802.11e HCF with Scheduled Access can do that sort of thing, but pure contention networks, such as 802.11e with EDCF can’t.

    To my knowlwedge, DOCSIS doesn’t have a poller; do you know otherwise?

    You should also be advised that traffic shaping BitTorrent traffic isn’t something unique to Comcast. Dozens of cable-based ISPs around the world do it today and have done it for a long time. It’s normal ISP behavior to throttle BT, and they’d be remiss not to.

  • http://bennett.com/blog Richard Bennett

    As a follow-up, wirelessman, I suggest you read “The Interaction Between the DOCSIS 1.1/2.0 MAC Protocol and TCP Application Performance”:

    “We have developed a model of the Data over Cable (DOCSIS) 1.1/2.0 MAC and physical layers using
    the ‘ns’ simulation package [2]. In previous work, we reported on the impact of several DOCSIS
    operating parameters on TCP/IP performance [3]. In this paper we extend those results by looking in
    greater detail at the impact that the MAC layer has on TCP performance when using the DOCSIS best
    effort service. We show that the interaction between DOCSIS and TCP exposes a denial of service
    vulnerability. By taking advantage of the inefficiency surrounding upstream transmissions, a hacker can severely impact network performance.”

    In effect, several BT streams in the DOCSIS return path mimics a DoS attack to non-BT users. That’s not cool.

  • http://bennett.com/blog Richard Bennett

    Wirelessman, you’re assuming that the approach to traffic scheduling that you’re familiar with is available on DOCSIS networks. Polled networks, such as 802.11e HCF with Scheduled Access can do that sort of thing, but pure contention networks, such as 802.11e with EDCF can’t.

    To my knowlwedge, DOCSIS doesn’t have a poller; do you know otherwise?

    You should also be advised that traffic shaping BitTorrent traffic isn’t something unique to Comcast. Dozens of cable-based ISPs around the world do it today and have done it for a long time. It’s normal ISP behavior to throttle BT, and they’d be remiss not to.

  • http://bennett.com/blog Richard Bennett

    As a follow-up, wirelessman, I suggest you read “The Interaction Between the DOCSIS 1.1/2.0 MAC Protocol and TCP Application Performance”:

    “We have developed a model of the Data over Cable (DOCSIS) 1.1/2.0 MAC and physical layers using
    the ‘ns’ simulation package [2]. In previous work, we reported on the impact of several DOCSIS
    operating parameters on TCP/IP performance [3]. In this paper we extend those results by looking in
    greater detail at the impact that the MAC layer has on TCP performance when using the DOCSIS best
    effort service. We show that the interaction between DOCSIS and TCP exposes a denial of service
    vulnerability. By taking advantage of the inefficiency surrounding upstream transmissions, a hacker can severely impact network performance.”

    In effect, several BT streams in the DOCSIS return path mimics a DoS attack to non-BT users. That’s not cool.

  • Ryan Radia

    Richard, why doesn’t Comcast use application layer packet shaping to deprioritize torrent traffic if it’s a network drain? Rogers and Shaw don’t seem to have a problem throttling Bittorrent traffic without forging RST packets. They have the same asymmetric cable network architecture as Comcast along with the limitations it imposes.

    Comcast clearly has the ability to monitor how much bandwidth its users consume, as evidenced by its termination of high usage customers who exceed an undisclosed limit. Charging extra to users who generate excessive network traffic is a much better idea than the one-size-fits-all Sandvine tool, which is terribly imprecise and goes against the nature of TCP connections. Comcast won’t even admit what it’s doing, claiming if it discloses its methods they will be easier to circumvent.

    I am certainly not calling for more regulation, and Comcast has the right to manage its network as it wishes. But the negative press and massive user outcry are well-deserved, and hopefully market pressures result in Comcast either changing its policy or becoming more transparent in how it manages its network. Considering Comcast has a government-granted monopoly in many areas, it’s particularly harmful that they are stopping users from seeding considering how many people have no alternatives available to them.

    It seems Comcast has inadvertently given the net neutrality proponents a lot of ammunition by choosing Sandvine to limit torrenting. I’m scratching my head as to why Comcast execs would resort to this technology, especially given the last thing they want is net neutrality laws. They just made it a whole lot easier for advocates of regulation to make their case using a specific instance of a company manipulating internet traffic.

    Comparing P2P to spam is disingenuous. Users who pay for service want to download and upload using bittorrent, unlike hijacked machines which use bandwidth that no paying customers care about. There are lots of legitimate uses for peer to peer nowadays and it is no longer used solely by a tiny slice of the internet population. Comcast is selling internet access, and seeding torrent files is included unless specified otherwise. If Comcast is going to prevent its customers from engaging in a legitimate, commonplace internet activity it should be straightforward about it instead of using opaque language and sidestepping the issue.

  • http://bennett.com/blog Richard Bennett

    Comcast is very straightforward about disallowing servers in its network, and BT seeds are nothing but servers.

    Comcast’s sin is simply that they’ve tried to make money from cables in the ground, so no matter what network management technique they use, some demagogue will accuse them of identity theft or something worse.

    Network Neutrality isn’t driven by abuse, it’s driven by ignorance. God forbid we ever see a real case of abuse, the critics won’t be credible.

  • Ryan Radia

    Richard, why doesn’t Comcast use application layer packet shaping to deprioritize torrent traffic if it’s a network drain? Rogers and Shaw don’t seem to have a problem throttling Bittorrent traffic without forging RST packets. They have the same asymmetric cable network architecture as Comcast along with the limitations it imposes.

    Comcast clearly has the ability to monitor how much bandwidth its users consume, as evidenced by its termination of high usage customers who exceed an undisclosed limit. Charging extra to users who generate excessive network traffic is a much better idea than the one-size-fits-all Sandvine tool, which is terribly imprecise and goes against the nature of TCP connections. Comcast won’t even admit what it’s doing, claiming if it discloses its methods they will be easier to circumvent.

    I am certainly not calling for more regulation, and Comcast has the right to manage its network as it wishes. But the negative press and massive user outcry are well-deserved, and hopefully market pressures result in Comcast either changing its policy or becoming more transparent in how it manages its network. Considering Comcast has a government-granted monopoly in many areas, it’s particularly harmful that they are stopping users from seeding considering how many people have no alternatives available to them.

    It seems Comcast has inadvertently given the net neutrality proponents a lot of ammunition by choosing Sandvine to limit torrenting. I’m scratching my head as to why Comcast execs would resort to this technology, especially given the last thing they want is net neutrality laws. They just made it a whole lot easier for advocates of regulation to make their case using a specific instance of a company manipulating internet traffic.

    Comparing P2P to spam is disingenuous. Users who pay for service want to download and upload using bittorrent, unlike hijacked machines which use bandwidth that no paying customers care about. There are lots of legitimate uses for peer to peer nowadays and it is no longer used solely by a tiny slice of the internet population. Comcast is selling internet access, and seeding torrent files is included unless specified otherwise. If Comcast is going to prevent its customers from engaging in a legitimate, commonplace internet activity it should be straightforward about it instead of using opaque language and sidestepping the issue.

  • http://bennett.com/blog Richard Bennett

    Comcast is very straightforward about disallowing servers in its network, and BT seeds are nothing but servers.

    Comcast’s sin is simply that they’ve tried to make money from cables in the ground, so no matter what network management technique they use, some demagogue will accuse them of identity theft or something worse.

    Network Neutrality isn’t driven by abuse, it’s driven by ignorance. God forbid we ever see a real case of abuse, the critics won’t be credible.

  • http://enigmafoundry.wordpress.com enigma_foundry

    Richard Bennett says:

    The essence of Tom’s argument seems to be that TCP Resets are bad, and dropping TCP segments is good (that being the only other known way of “throttling” TCP traffic.)

    Richard, this is a complete deliberate misconstruing of Tom’s point, which is very, very simple:

    I’m not trying to imply that Comcast should allow unrestricted BT traffic. But they should disclose what they’re doing, and use less destructive means of managing their network.

    RICHARD: Tom is just Saying: Comcast is lying, and that is wrong.

    I hope Comcast gets a huge class action lawsuit and loses tons of money, because they have committed FRAUD.

    Here is Comcasts own FAQ today, November 16, 2007:

    FAQs / Product Information / Comcast High-Speed Internet / Connection

    Do you block access to peer-to-peer applications like BitTorrent?

    No. We do not block access to any Web site or applications, including BitTorrent. Our customers use the Internet for downloading and uploading files, watching movies and videos, streaming music, sharing digital photos, accessing numerous peer-to-peer sites, VOIP applications like Vonage, and thousands of other applications online.

  • http://enigmafoundry.wordpress.com enigma_foundry

    Richard Bennett says:

    The essence of Tom’s argument seems to be that TCP Resets are bad, and dropping TCP segments is good (that being the only other known way of “throttling” TCP traffic.)

    Richard, this is a complete deliberate misconstruing of Tom’s point, which is very, very simple:

    I’m not trying to imply that Comcast should allow unrestricted BT traffic. But they should disclose what they’re doing, and use less destructive means of managing their network.

    RICHARD: Tom is just Saying: Comcast is lying, and that is wrong.

    I hope Comcast gets a huge class action lawsuit and loses tons of money, because they have committed FRAUD.

    Here is Comcasts own FAQ today, November 16, 2007:

    FAQs / Product Information / Comcast High-Speed Internet / Connection

    Do you block access to peer-to-peer applications like BitTorrent?

    No. We do not block access to any Web site or applications, including BitTorrent. Our customers use the Internet for downloading and uploading files, watching movies and videos, streaming music, sharing digital photos, accessing numerous peer-to-peer sites, VOIP applications like Vonage, and thousands of other applications online.

  • http://enigmafoundry.wordpress.com enigma_foundry

    Comcast is very straightforward about disallowing servers in its network, and BT seeds are nothing but servers.
    Let’s parse this post by richard Bennett:

    Comcast’s sin is simply that they’ve tried to make money from cables in the ground, so no matter what network management technique they use, some demagogue will accuse them of identity theft or something worse.

    No, the sin is that they have lied to their customers, and were deceitful and evasive when asked about their behavior. Why did they lie about their behavior? because they KNEW IT WAS WRONG.

    Network Neutrality isn’t driven by abuse, it’s driven by ignorance. God forbid we ever see a real case of abuse, the critics won’t be credible.

    Deliberate ignorance is the only excuse for not knowing of very credible examples of violations of network neutrality. I have listed many here:

    http://enigmafoundry.wordpress.com/2007/08/14/we-dont-need-no-thought-control/

    To make no mention of the great firewall of China.

  • http://enigmafoundry.wordpress.com enigma_foundry

    Comcast is very straightforward about disallowing servers in its network, and BT seeds are nothing but servers.
    Let’s parse this post by richard Bennett:

    Comcast’s sin is simply that they’ve tried to make money from cables in the ground, so no matter what network management technique they use, some demagogue will accuse them of identity theft or something worse.

    No, the sin is that they have lied to their customers, and were deceitful and evasive when asked about their behavior. Why did they lie about their behavior? because they KNEW IT WAS WRONG.

    Network Neutrality isn’t driven by abuse, it’s driven by ignorance. God forbid we ever see a real case of abuse, the critics won’t be credible.

    Deliberate ignorance is the only excuse for not knowing of very credible examples of violations of network neutrality. I have listed many here:

    http://enigmafoundry.wordpress.com/2007/08/14/w

    To make no mention of the great firewall of China.

  • Jim Oase

    Richard, I think you are taking your eye off the target. Comcast needs to say what they mean and mean what they say. To do otherwise is unethical at minimum.

    If they choose to use slippery language to describe their product and the customer buys it, shame on the customer. To use the language and deny using it because it is within the law is the sign of a bottom feeder. Our laws describe minimum acceptable behavior. When someone say within the law they are often describing their kind of minimum behavior….that of a bottom feeder.

    Comcast is not a person, a person made the decision to have Comcast’s product work at minimum standards. A person gave the directions to describe Comcast’s products in elusive ways. That person needs to be credited as appropriate. People who work for companies with questionable ethics are saying their ethical standards are for sale.

    The technical issues are interesting. At the end of the day its ethics that builds a future. People tend to do what they have done before. Bottom feeders have fed on the bottom before.

    Jim

  • Jim Oase

    Richard, I think you are taking your eye off the target. Comcast needs to say what they mean and mean what they say. To do otherwise is unethical at minimum.

    If they choose to use slippery language to describe their product and the customer buys it, shame on the customer. To use the language and deny using it because it is within the law is the sign of a bottom feeder. Our laws describe minimum acceptable behavior. When someone say within the law they are often describing their kind of minimum behavior….that of a bottom feeder.

    Comcast is not a person, a person made the decision to have Comcast’s product work at minimum standards. A person gave the directions to describe Comcast’s products in elusive ways. That person needs to be credited as appropriate. People who work for companies with questionable ethics are saying their ethical standards are for sale.

    The technical issues are interesting. At the end of the day its ethics that builds a future. People tend to do what they have done before. Bottom feeders have fed on the bottom before.

    Jim

  • Pingback: Camfrog Room

  • Pingback: devenir rentier

  • Pingback: The Tao of Badass

  • Pingback: premier league singapore

  • Pingback: anti wrinkle serum

Previous post:

Next post: