Day seven of the CRTC's network management hearing featured just one company: Bell. As the prime target for much of the criticism associated with traffic management, Bell executives faced questions for nearly three hours, far longer than anyone else.
Key points included new details on Bell's traffic and traffic management practices, claims that the company cannot separate retail and wholesale Internet traffic, and the company's support for a "reasonableness" standard, rather than the "least intrusive" approach advocated by several groups.
Today's summary was again compiled by Sean Murtha, a law student at the University of Ottawa. Other coverage available from the National Post liveblog, CBC.ca, the National Post, and twitter feeds from CIPPIC and me.
CRTC Net Neutrality Hearings – Day 7, Bell – Tuesday July 14th, 2009
Bell Aliant Regional Communications, Limited Partnership, and Bell Canada
Appearing for Bell is Mr. Jonathan Daniels, VP for Regulatory Law, and others. In its opening remarks, Bell Canada stated that “at the heart of this hearing is a dispute about innovation. Bell argued that some parties have argued, without any proof, "that the threat posed by application specific shaping is so severe, that it must be stopped. Bell posited that where some see the need to impose regulatory rules to protect application innovation, they see their demands as seeking regulatory rules that "would limit, or even ban, innovation at the network level." Bell stated that "imposing ex ante rules that limit or ban innovation at any layer is by definition a greater threat to innovation than having no ex ante rules at all.
Bell put forward a 3 pronged approach to effectively address network congestion:
- Capital Spending – Bell stated that they continue to invest in their network, but noted that "unlimited capital spending is not a luxury the companies’ enjoy". The economic and market realities of high-speed internet access clearly show that a pure capital spend solution not a sustainable model, and they argued that it is too simplistic to assume that Bell can build their way out of the problem.
- Pricing and New Business Models – In today’s market, prices are not granular enough to eliminate congestion during peak hours. Prices motivate users to reduce consumption over the month and deal with average usage. "While pricing helps, it is not the sole solution".
- Bandwidth Management – Bell stated that they "firmly believe that our traffic shaping practices are acceptable". They are designed to improve the overall experience for the vast majority of users. They noted that the practices they implement are (i) narrowly focused and limited to non-time sensitive traffic, (ii) confined to a defined period of time; and (iii) do not block access to content.
In regards to privacy, Bell confirmed that their DPI does not, and cannot, inspect the user content of communications, and guaranteed that "the DPI technology in our network is being used only for purposes of traffic management". It does not inspect user content, and it is not being used for marketing purposes.
For wholesale, Bell stated that "two assertions must be corrected". First, contrary to what many have told the CRTC, retail users and wholesale GAS users share the same network. As a result of this, wholesale traffic impacts retail traffic, and vice-versa. Secondly, Bell stated that "wholesale ISPs have attempted to dismiss their impact on our retail network by pointing to their overall market share." They argued, however, that the reality is that wholesale traffic is significant, as in the month of May 2009, wholesale GAS traffic accounted for 31% or their total GAS and retail traffic, and GAS traffic accounted for 36% of their total P2P traffic.
Bell also noted that some want to adopt an "Oakes-type" test, including the least intrusive standard. Bell argued that the appropriate test for ITMPs is not one which has the regulator substituting its opinion for that of market participants. Bell submitted "that the appropriate test for reviewing an ITMP (and the one required by the statute) is a reasonableness test, not a least intrusive test".
Bell also proposed the following 3 guidelines, which they developed for the CRTC’s consideration:
- Limiting Negative Impacts – ISPs should make reasonable efforts to limit the negative impacts of ITMPs on users, services, protocols or applications. Implicit in what is "reasonable" is the recognition that different networks face different problems and therefore ISPs require the flexibility to reflect those distinctions in adopting ITMPs.
- Transparency – ISPs should disclose to affected retail and wholesale customers "the general nature of implemented ITMPs in a manner that does not compromise the security of networks and commercially sensitive information.
- Privacy – ISPs should implement ITMPs in a manner consistent with applicable privacy laws. As an example, Bell mentioned data used for the purpose of ITMPs should not be used for any other purpose without customer consent.
Bell stated that they felt that s. 36 of the Telecommunications Act will rarely be applicable to issues that arise with respect to ITMPs. To the extent they do arise, Bell said it is possible to address them under the first guideline which provides as an example that ISPs should not implement ITMPs that block content based on criteria such as its source, ownership or destination unless such activity is necessary to protect the integrity of the Bell network. Bell concluded by stating that "imposing rules that limit or ban innovation at any layer is by definition a greater threat to innovation than having no rules at all".
Chairman: P2P seems to be a flashpoint in these hearings. We have heard that it is hard to detect what traffic is actually P2P, because some P2P traffic is encrypted. What is your view of this, because you are obviously doing P2P throttling?
Bell: The short answer is the traffic shaping Bell does is sufficiently effective to achieve our objectives, as we wouldn’t be doing it if we did not think it was effective. Bell has a very accurate ability to recognize P2P, the evidence of that is that, overall, 27% of our traffic is P2P, but at peak hours that is limited to 14% because of our traffic shaping.
Chairman: Do you do any ex post facto analysis to see if you have missed any P2P?
Bell: We know some P2P gets through.
Chairman: Is the P2P traffic from wholesale customers relatively proportional within your network?
Bell: There is a slight difference, but in terms of overall, it is roughly the same (i.e. it is not a major concern).
Chairman: VT said yesterday that they only use commercial pricing mechanisms to deal with congestion issues, Shaw monitors and throttles beyond a certain threshold, Bell throttles during peak hours, Cogeco throttles 24/7. Why are there so many approaches taken by the different ISPs?
Bell: The short answer is that each ISP looks at it from the perspective of what works on their network. We look at solutions that address needs in our network, and the solutions that we saw that would address it we followed. Flexibility is important. That is why the CRTC should not dictate rules, because we are not aware of any unreasonable practices in the Canadian market today.
Chairman: Why can you not use the Shaw approach, which to me appears to be very appropriate. As you approach the dangerous levels, that is when you throttle. In your opening remarks, you have advocated for a "reasonableness" approach – wouldn’t it be reasonable to only throttle when it is needed?
Bell: You have to keep in mind that there is a big technical difference between cable networks and telephone networks. The CMTS system makes a difference. The equivalent in Bell is DSLAM. On top of that, Bell’s network is "highly meshed", and congestion occurs in the aggregation network. Bell has 1500 links in that network. You cannot manage on an individual link basis in a network like that. This is a technological problem.
Chairman: So even if you wanted to, you couldn’t act like Shaw does?
Chairman: What stops you from making your pricing more granular?
Bell: As we pointed out, the pricing mechanism is not granular enough to deal with peak period problems. To do this, we would have to have pricing based on peak periods. Sandvine discussed this issue earlier this week.
Chairman: What is the difference between internet use and mobile phone use, because mobile phone plans have different prices for different times of the day. Why can’t the same be done for the internet?
Bell: It has to do with consumer education. Mobile phone consumers have been educated for years about mobile phone plans, and peak period rates etc. I am not saying it is not possible, but we would have to educate the public. However, this would likely lead to price increases.
Chairman: Are you afraid of being a market leader in this area, and are you afraid of losing customers?
Bell: The market is not there yet. If the market moves there, then we can follow. However, it would require a greater customer understanding of things like bandwidth levels. We come back to the notion of what is the problem we are trying to address? Most customers are very happy with their internet services (Decima-Harris study, 87% of users are happy with their internet services). The trade off with doing to a peak period usage billing system would add a layer of complexity which would undermine the use of the internet.
Chairman: Rogers said that economic disciplines only go so far, because that only addresses the downstream, when upstream is also an issue. What is your view? Would introducing economic incentives not address the upstream?
Bell: The difference between us and Rogers is that their problem is solely on the upstream, while Bell, with DSL, is more on the downstream. Bell could fashion a "counter" to allocate more for upstream or downstream. However, for economic discipline to be successful, the cap would have to be so low as to be punitive to the customers.
Chairman: VT talked yesterday about using only economic methods to deal with problems. If you throttle uploading, you also create an effect on the downloading.
Bell: If you are using the internet, you will use both the upstream and downstream, because it is dynamic.
Chairman: If you have network problems, do you pass them down to your wholesale customers?
Bell: Anticipating this question, in our opening statement we have attached a diagram. If you buy our retail service, or our GAS service, you will go through our aggregation network. If an ISP wants to buy GAS service, they can interconnect at one point in the Bell network, at a Bell CO in the geographic area they are located. At that time, the ISP then has access to all consumers that Bell’s technology can reach. When the traffic hits the BAS, the BAS can distinguish between Bell’s retail and wholesale traffic.
ISPs pay Bell for each end user they sign up, and also pay a monthly fee for the total amount of bandwidth they need. They are really paying a fee for a port, and that fee represents less than 2% of their payment for GAS. ISPs are not actually paying for the bandwidth itself, they are "only paying for access to the door". So, how is it that they are not segregated (i.e. retail from wholesale)? The traffic is not segregated in the aggregation network, and that is where the congestion problem is. However, ISPs do have the option of paying a higher price to skip the DPI, and they can also skip the traffic shaping.
Bell can clearly turn on traffic shaping on for retail customers, and off for wholesale customers. A problem is that Bell cannot turn their traffic shaping on for only 1 wholesale customer. They have to either turn it on for all wholesale customers, or no wholesale customers. Therefore, for those ISPs that go through the aggregation network, Bell’s traffic shaping applies.There is no way, technically, in the protocols to recognize the difference between the different wholesale customers.
Chairman: So your problem with congestion exists in Bell’s aggregation network?
Chairman: What would happen if the CRTC said that throttling was a violation, and that you were not allowed to do it?
Bell: I think we have showed that P2P is a problem during peak periods, so removal of shaping would lead to congestion, and that in turn would lead to poor customer service. Then, I could do 2 things: 1) I could do nothing and let customers suffer, which I would not do, or 2) Bell would have to invest more, which would mean that we would have to redirect capital within Bell, and raise prices for consumers.
Chairman: Yesterday we heard from UdC, and a speaker said that there should be clear evidence that there is congestion in the network, and he said that there is little evidence that this congestion actually does exist. Can you, Bell, show me information that throttled applications cause significant congestion?
Bell: The answer is yes. P2P represents over 27% of traffic on our network. When people say it causes congestion, it is part of the problem for congestion, but is not the only problem. The reason that we focus on it is because it causes congestion, but also because of how it uniquely acts in a congested state. It is there that P2P takes a disproportionate amount of bandwidth.
Chairman: I would understand if you do not want to make this data public but if you do, that is all the better.
Bell: I think that this information is already in the public record, so we will be happy to take you through it. The first part of the 3 part answer is that there is congestion in our network (page 32 of Bell’s July 8th 2009 submission, figure 12.
Chairman: I do not have that submission in front of me. Why don’t you do it in a written format.
Bell: There are some things that would be better explained orally. In reference to yesterday’s discussion of shallow packet inspection (SPI), this is what struck me, as a lawyer. If you are doing SPI to make distinctions, you are trying to recognize what applications involve. To say don’t do application specific, and use SPI, does not make any sense. SPI is used to make distinctions based on applications. The expert witness’ concluding remark was that there was no current commercial use of SPI. We agree with that, as there is no SPI tool on the market right now.
Bell then showed another graph, attached to their submission. It endeavoured to show that P2P takes an unfair, disproportionate amount of space in a congested state. In a congested state, P2P is not as affected as users who are not using P2P. By limiting only file sharing (as we do not touch time sensitive applications), we can increase overall user experience and the people using it are relatively unaffected. Traffic shaping is necessary to guarantee user satisfaction.
Chairman: You seem to be the only ones who traffic shape the downloading as well as the uploading.
Bell: In Canada. There are many international carriers who traffic shape both upstream and downstream.
Chairman: The reason that you throttle downloads, as well as uploads, is technical then? It is a technological requirement for you?
Bell Canada: Yes.
Chairman: I have changed my mind, can you take the panel through the document that you referenced earlier?
Bell: Our data shows escalating congestion. Queues become overloaded, packets or cells are dropped, and require re-transmission, either by TCP or in the case of UDP, they are lost altogether. If you are using TCP, the application will be asked to retransmit (a good example is a VOIP call). Those packets are lost forever, so they are repeated.
So, P2P is a major cause of congestion. Every packet that is going through a congested device is a contributor to that congestion. In our case, 27% of our traffic is P2P. So, why do we handle it the way we do? Most of the traffic on the network assumes some immediacy. The issue is how do you handle customer experience at the peak? What we do is prioritize traffic which has immediacy (such as time sensitive applications) over machine to machine transfers.
Chairman: Do you use utilization data?
Bell: All companies use utilization data, we look at queues in the devices, and it is a reasonable proxy. Utilization is a reasonable proxy for large amounts of utilization data? Remember, we do not have a problem with P2P. Our top 10% of users are responsible for 96% of P2P, amongst retail users. Overall, we believe that this is a user-friendly way to do it, and users use P2P to increase their use of other functions.
Chairman: Switching gears, you have said that aggregated data is not being used for marketing purposes, and it should not be used for anything else without customer consent. This is weak, as you are not allowed to use it for other purposes.
Bell: What we are using DPI is for traffic shaping only. Privacy legislation (PIPEDA) allows us to use it for other purposes, if we went back to customers and got consent. We recognize that people are concerned about privacy.
Chairman: PIPIDEA would allow you to use data for marketing, if you had customer consent?
Bell: Marketing is absolutely not what we use it for.
Chairman: That was not my question. Are you telling me you could use it for marketing purposes if you had customer consent? Is that what the privacy legislation allows?
Bell: That is what PIPEDA is for – it is based on customer consent, but currently we do not use it for marketing.
Chairman: I am trying to understand why you are so focused on a "reasonableness test", instead of a "least intrusive test"? Shouldn’t traffic management be least intrusive, as it is consistent with your philosophy of acting for the benefit of your customers?
Bell: The problem I have with "least intrusive test" is that, ultimately, there has to be a judgement made as to the one correct least intrusive method. Ultimately, if the test is "least intrusive", what is going to happen is many different people are going to come forward with suggestions as to what is "least intrusive". Ultimately, the CRTC will have to decide what is least intrusive, and would force ISPs to act in a certain way.
Chairman: We have talked about essential services. If you have to determine what is least intrusive, it would be on you to determine what you could do. It would be a reasonable decision based on evidence brought before the Commission.
Bell: For essential services, there were actual determinations made. The best way to think about this is the way that a judge looks at an administrative tribunal decision. There is a difference between a correctness standard and a reasonableness standard.
Chairman: This would be ex post, not ex ante. Ex post would alleviate some fears for you, certainly?
Bell: Correct. We are strongly opposed to ex ante. The problem with protocol agnostic focuses on people who are using applications, and it will slow down all of a user’s traffic, even time-sensitive traffic. Some will say this is good, because it does not differentiate between applications. We do not think this is unreasonable, but it is not what we have done. The OIC, and others, will say this is less intrusive than our method. We disagree; we think our way of shaping applications is less intrusive. Ultimately the CRTC would have to decide.
Katz: First, your proposal is directed at the public internet. Do your networks co-exist, or are they independent?
Bell: Our networks are all "in the same pipes".
Katz: So if your managed IP has a high incidence of traffic,that could potentially congest the public internet?
Bell: Let me give you a broader answer. In our network there are many traffic types, some have a higher priority throughout the network. If we did not engineer the network correctly, there could be congestion. If you are going to offer IPTV, you would not under-build, you will make sure to have the capacity.
Katz: Telus said last week that a company would not want to negatively impact the public internet. Do you agree with their views?
Bell: I do. With IPTV architecture, you do not impact the capacity of the public internet.
Bell: Remember, as mentioned, we cannot, in our network, distinguish between ISPs, and have different rules. We have to, in the aggregation network, treat everyone the same.
Katz: Primus does use DSL. They claim that they can traffic manage and do some of the things you said you cannot do. Are your networks the same?
Bell: I do not think they said that they cannot do something that we cannot. They said, not when using our GAS service, they make a distinction between time sensitive data (i.e. VOIP and Gaming) and that which is not.
Katz: But Primus said that they only do shaping during congestion periods. Your traffic shaping is not focused at all (it’s on for 10 hours a day). They then alleviate all their shaping tools when they are not required.
Bell: I am not aware of signalling mechanisms in the DSL network which allows one to identify real time in any automated fashion.
Katz: However, you have said your methods are "narrowly focused". How is this the case when you have traffic shaping on so often (as I mentioned, 10 hours a day)?
Bell: The first aspect is that we make a distinction when it is close to peak periods. The key aspect is that it is limited to only file sharing, non-time sensitive P2P traffic. There are many reasons why people face problems, but we have not had collateral damage.
Katz: Is this traffic shaping method (i.e. throttling P2P) available for public consumption? Do you have information as to your traffic management available to the public?
Bell: We do have some information on traffic shaping methods, but we have not had the speeds we have throttled down to. We are adding those speeds onto the website though, soon.
Katz: So 31% of wholesale customers use GAS?
Bell: No, 15% of end users and GAS end users, but they represent 31% of traffic. While this is disproportionate, this is not generally a problem, only during periods of congestion.
Katz: For those customers that want, they can pass the "best efforts basis"?
Bell: No, they will pass our traffic shaping.
Katz: How many of your wholesale customers purchase non-traffic shaped services, on dedicated lines?
Bell: The number is low.
Katz: One of the reasons why the tariff is more expensive is because you are dedicating certain things to them?
Bell: No, it is "nailed up" right through to the aggregated high speed base.
Molnar: Since we were talking about GAS, I have a similar question. In our conversation with Execulink earlier in the week, we talked about 2 components. I would like to understand, based on an example from Execulink where they said they can provide each customer with up to 11mbps, and their bandwidth is 100g, when you throttle on the customer. What impact does this have?
Bell: Certainly by lowering the P2P, if their customers are not going to do things with that extra bandwidth, it does not have an impact. If their end users are mainly P2P users, then our practices could impact them. What I can tell you is that, talking to our wholesale team, we bought a certain about of bandwidth (said the ISPs), and we have seen no correlation between our traffic shaping and reduction requests. We have seen the same typical growth.
Molnar: You said you saw no difference after traffic shaping, with the amount of bandwidth as being transported?
Bell: I do not know about that. Look at it from an ISP perspective. They say they need 1 pipe to connect to you whole network, and at that point, the ISP "chooses the door size" they will have.
Molnar: Is there anything within the tariff that would prohibit them (an ISP) from moving down to a lower size, and getting a different capacity?
Bell: No, and factually, we have not seen it. ISPs are signing up new end users, and want "more doors". That is not to say that individual ISPs have not gone down, but we are talking trends. For the most part, ISPs have not wanted to move down to a smaller door.
Molnar: But ISPs can change their "orders", and subsequently change their costs? This is possible for them to do?
Bell: Yes, I believe that is the case. The reason I focus on a "door" instead of a "pipe", it is priced only at the point they connect, which gives them access to our whole network province wide. They do purchase a "pipe" within our network, the purchase a "door" to access our entire network, within a province.
Molnar: You said that you want guidelines which are flexible enough. Would you say DPI is a "network innovation"?
Bell: Absolutely. DPI is a network innovation which allows us to improve the user experience. Another example is our ability to block spam on our networks. We believe our practice is an innovation, and it is important.
Molnar: Juniper said that congestion management could occur at the customer level. Instead of being the choice of a network provider, the choice could be with the customer. They did mention this does not currently happen in Canada, but have you looked at these types of options? Is this part of the evolutionary plan, that these controls could one day be held by the customer? If you have to decide between the marketing person and the engineer as to who is going to answer, I think you already have answered it.
Bell: There are many different reasonable solutions out there, although they are not all currently reasonable. From a white-listing point of view, it has potential. I don’t think this would be viable in today’s market, and it is not something we will implement anytime soon. A network of our size, we have a very good track record of breaking things that do not scale well. While that is an interesting model, I would like to see if go further in the market first. We are not going to rule out anything. We will have to decide in the future whether that is the best way to go.
Lamarre: I want to try to answer one question you have asked us. At the start of your presentation, you said at the heart of this hearing is a dispute about innovation. While that is true, that is not all this hearing is about. You have asked why traffic management practices are an issue if users are happy. However, we have to be concerned about all users, even those that fall within the minority. That is why we are inquiring, and that is why we are being picky, if that is how you think we are acting. You have also noted that some presenters at these hearings have made allegations, but without proof or data to support these allegations. The problem is that the data is held by ISPs, and is not always available to the public. Is there any way that from your perspective, you can demonstrate that there is actually no problem?
Bell: Us being Bell, when we do something wrong, we hear it from customers, who can speak with their feet and will leave us. We do market research, and nothing has pointed to our P2P shaping as being a problem with consumers.
Lamarre: Let’s talk about consumer experience and price. I still do not understand why you throttle P2P on both the upload and download. Rogers throttles only upload, not download, and this is done because customer experience is important as far as download is concerned. I understand that Rogers is cable and you are DSP. TELUS is also DSL, and they do not throttle at all. So, this system that Bell uses does not seem to be a good customer tailored approach, does it?
Bell: I cannot speak as to why TELUS does not. If you had this hearing years ago, you would have asked Rogers why they do not, and Bell does. In the Atlantic Provinces, we do not traffic shape. In the central provinces, we have seen congestion problems that we have not seen in the Atlantic Provinces. There is not the same level of peak period use the Atlantic Provinces. Also, the network is designed differently in Atlantic Provinces. This does not mean that we will not traffic shape if congestion becomes an issue, but currently we do not find it necessary.
Lamarre: You have also put into place economic measures, reducing the amount of bandwidth by 10 times (20g to 2g per month). However, the prices are almost the same as they were before, if not the same. Is this not enough of a measure, or do you still need to throttle? By offering different packages, is this not enough?
Bell: For us, pricing does not seem to be enough. While we feel that it is important (it is part of our 3 pronged approach), we feel the other steps we outlined are also necessary.
Lamarre: The guidelines you are proposing, would this be enough for traffic shaping on wireless, mobile wireless?
Bell: Absolutely. In limiting the negative impacts, different networks face different problems. What is reasonable in a wireless network may be very different from what is reasonable in a wireline network. Flexibility is important.
Lamarre: Should we make a distinction between fixed wireless and mobile wireless?
Bell; If it is justified, it is part of the test. If it is determined it makes sense to make a distinction, then yes.
Lamarre: On the one hand, PIPEDA is a law that is general in nature. When the act was passed, internet traffic was not a major issue. The public commissionaire raised the point that the CRTC’s role under the Privacy Act is that the CRTC has jurisdiction on privacy over communications networks. Under the Telecommunications Act, we also have jurisdiction over privacy issues.
Bell: We understand that the CRTC has a mandate to protect privacy. We are also subject to PIPEDA. It is our responsibility to ensure that our organization meets obligations under PIPEDA, and other rules that flow from the CRTC. Under the PIPEDA system we received some complaints, but the complaints always came through the OPC. We now have a complaint filed by CIPPIC, and in collaboration with the OPC to settle the issue.
Lamarre; Question about DPI vs. SPI. The question raised by the UdC yesterday about privacy protection makes me ask, which of these is more invasive?
Bell: As far as privacy issues, we do not have any information relating to the packet to which the technology applies. We do not keep the information.
Lamarre: Even if you do not keep it, I think there are still privacy issues. There are some worries on the part of users.
Bell: We apply the necessary policies for traffic shaping, then the process is over. The intent of SPI is identical to DPI.
Lamarre: I know the intent is the same. My question is, with DPI, do you get more detailed information?
Bell: There is a little more information available with DPI.
Lamarre: You also said you were not aware of any SPI technology that is currently on the market. However, you must consider that it may be difficult to get something on the market if there is no demand. So, does Bell plan on testing SPI technology, to see if it could be as accurate as DPI?
Bell: We feel confident that SPI is not as accurate.
Lamarre: In your guidelines you suggest, instead of an Oakes-type test, a reasonableness test. I would state the Oakes test to ensure that citizens are protected, and have their fundamental rights protected. Don’t you think that we should adopt an Oakes-type test when it comes to privacy rights?
Bell: It is a big difference, because the Charter protects rights infringed by the government, not private companies.
Lamarre: But the CRTC had the duty to protect privacy, and the CRTC is part of the government.
Bell: I agree, but there is a fundamental difference between how the government makes rules, and how companies operate. Customers have options, and can change companies. The statute provides for a reasonableness test, and we do not think that a constitutional like test which applies to the government, which is the ultimate monopoly, should be the standard for private actors.
Lamarre: Not even to protect a fundamental right like privacy?
Bell: I do not want to downplay privacy, as privacy is very important for us.
Denton: You mentioned that you were approaching a more "granular approach" to bandwidth management. Can you elaborate?
Bell: We have some stuff in the lab, and we are looking at the "Cox Solution" (in the U.S., they are trialing about time-sensitive vs. non-time sensitive distinction).
Denton: You have dropped your bit-cap by an order of magnitude. Was there a reason?
Bell: We looked at market trends, customer needs, competition. We made our decision based on what the market would bare.
Denton: That was a market oriented answer. Bell uses traffic management all the time?
Bell: Just to be clear, we do it from 4:30pm until 2am.
Denton: So you only do it at peak hours? Do those peak hours correlate to your periods of congestion?
Bell: Yes, those are our periods of congestion. However, that is not to say that an event, like Michael Jackson’s funeral, can cause congestion at a non-peak period time.
Denton: Is traffic management at the wholesale level reasonable?
Bell: It is reasonable. Bell cannot distinguish between wholesale providers.
Denton: Why is that? Is this just a software problem?
Bell: It is a protocol problem. The tunnel IDs are dynamic. At that level of the network, they are not permanently identified. That is the technical issue, but the other issue is, say Primus gives priority over its traffic, how does that work once it is aggregated in our network? If they try to shape it, it will conflict with our methods, and would not matter if they are giving priority in their networks. The only way we could leave it up to them would be if they did exactly the same thing we did, so we might as well do it for them.
Denton: So it seems to be a technical and economic reality.
Bell: Yes, and remember that we would have to make technical agreements with more than 100 companies, and they would have to end up following exactly the same practices that we already are doing for them.
Denton: Another major contention is that traffic management measures are an inadequacy of competition at the retail level. Do you have a comment on this?
Bell: People have pointed to England as the panacea. However, many ISPs there institute traffic shaping. Japan, which is considered a leader, in 2007 25% of ISPs traffic shaped (and this data is 2 years old). So, to make the connection that competition is the reason may not be the answer.
I also should state that the industry is highly competitive in Canada. There are many ISPs which differentiate themselves, and there are lots of options out there. We disagree with anyone who says there is not strong competition in the Canadian market.
Chairman: I think in regards to a "least intrusive test", you want to make sure that it is least intrusive not only on its affect, but for the ISPs themselves. The difference between that and "reasonableness" is not that great.
Bell: Our goal is the same in that regard. Our definition of reasonableness means "reasonable efforts to limit the negative impacts of ITMPs on users, services, protocols or applications".
Chairman: I do not think we are very far apart. Thank you very much, we spent more time with you, but you did start the whole thing.