CRTC Net neutrality Hearings – July 9, 2009
Canadian Association of Internet Providers (CAIP)
Summary:
- The issue before the table is competition on the Canadian ISP landscape; congestion is a symptom of the lack of competition.
- CRTC should prevent Dominant Carriers (DC) to stop ITMPs on their wholesale traffic, except for network security.
- In order to drive competition, the use of ITMPs (for retail) should be end-user determined, that is, consent to technologies must be required by the ISP.
- ITMPs that traffic control in an aggregate fashion, absent customer's consent, is contra s.36, 27(2) by interfering with the "purpose or meaning of telecommunications."
- Need to develop a uniform, quantifiable measure for congestion
Opening remarks:
Chris Tacit - Counsel
Tom Copeland – Chair of CAIP
CAIP brings to the table the perspective of independent ISPs on the commercial issues surrounding ITMPs. The role of the independent ISP industry is crucial for increasing competition and innovation. For example, VOIP was first adopted by independent ISPs. Independent ISPs are threatened by the market power of DCs: independent ISPs must obtain wholesale customers from the DCs while the DCs are the main competition with an overwhelming market share (95.5% of residential subscribers). This makes it difficult for independent ISPs to offer Internet services that are different from those of DCs. End-user consent to ITMPs is necessary for driving product differentiation and thus, competition, in the market. Building network capacity and unbundling networks are integral and less intrusive methods for network management. The assumption that unrestricted increases in Internet traffic can lead to "congestion" is unfounded and untested and undefined. Moreover, congestion should not drive the policy determination regarding ITMPs. ITMPs may be necessary for stopping spam, viruses and network attacks, but they should not be used on the wholesale level otherwise. Availability of diversity in service packages would be an indicator of healthy competition at the end-user level.
Responses to CRTC:
- ITMPs should be based on business model of the ISP, to drive competition
- ITMP disclosure: ISPs should disclose ITMPs to wholesale customers (but not the technical details), but these should only be for network attacks. Traffic management in retail services should be disclosed upon customer sign up.
- Privacy concerns: there is no need to use personal data in ITMPs although it may occur in the future for commercial reasons; there is also no need to go beyond the existing privacy framework (PIPEDA and CRTC decisions on confidentiality).
- Is application of ITMPs to wholesale services by retail ISPs appropriate? No, except for network attacks. DCs must refrain from applying ITMPs on wholesale customers on behalf of competitors, under s.27(2) and 36. If the retail ISP is going to make any modifications, it should be subject to wholesaler notification and also to end-users.
- The same rules should apply to all ISPs regardless of wireline or wireless
- Congestion should never drive policies relating to ITMPs.
- The situation warrants the development of uniform, quantifiable measures for congestion.
Questions:
Chair -
What do you mean by "Dominant Carrier"
CAIP -
The duopolistic nature of the market makes them jointly dominant; we don't think that CAIP would be here if it weren't for joint market dominance of cable companies and telcos.
Chair -
Are you suggesting that more unbundling will mean no more congestion?
CAIP -
Unbundling is a crucial part of the solution for the competitive industry.
Chair -
You submit that there should be no ITMPs as a wholesaler, which is tied in with your conjecture that we assume the presence of congestion. We are here because there is congestion, it's not an assumption. We are here to make sure that everybody can enjoy the Internet but that people who are capacity hogs don't hinder others.
CAIP -
Wholesale users have tiny fragment of the industry (95.5% come from cablecos and telcos) - when there's a congestion problem it's with their end-users. So, the wholesalers shouldn't suffer. Moreover, wholesale and retail traffic can be disaggregated, so there's no justification for retailers to control the traffic of a wholesaler who's internal traffic doesn't affect them.
Chair -
If congestion is the wrong lens to look at it, what is?
CAIP -
It's about consumer choice and competition. ITMPs can be used to provide for end-users bill management. When it hasn't been demonstrated that wholesale is causing any kind of congestion in the ILECs networks, and even if there were there would be other ways to deal with it.
Commissioner Katz -
How do your members differentiate their services from the DC where you are competing with them?
CAIP -
Many use the underlying infrastructure to develop private networks so the traffic never touches the public Internet, it just goes between computers internally (private LAN). Providers also have done VOIP protocol between two customers that never touches the pub Internet, just through the providers. Also our competition is in the fact that our wholesalers serve customers locally, but there's not much room to differentiate ourselves further b/c of the ITMPs.
Katz -
So a wireless retailer should be able to provide a better service than the incumbent who may have congestion practices?
CAIP -
The cost of consumption is the driving factor; so it's hard to draw that analogy just based on the market. Industry Canada has been focused on increasing spectrum. Existing competitors want to become masters of their own networks, by focusing on owning your own capacity.
Katz -
There's been evidence that if you look at international markets, they work better, partly because there is no distinction made between retail and wholesale. British Telecom (BT) is a monopoly in the UK, their retail arm is treated as all the other wholesalers, they've got to subscribe to it, but BT retail is treated no differently than resellers.
CAIP -
No one's saying there's no difference between retail and wholesale; the MTS evidence shows that in the UK they've unbundled a lot of the wholesale, so there's been a lot more competition and competition has made the net neutrality problem obsolete. CAIP is not in the reselling model of the UK. The CRTC shouldn't make its determination on ITMPs without the whole picture, we're trying to deal with the problem of net neutrality by looking at congestion - when it's about ensuring more competition in the marketplace. In the UK, the vertical integration of BT was causing problems with the expansion of the network. Since they've changed the model competition is vibrant and the unbundling model has been helping. There's no controversy because the marketplace is taking care of it, there won't be any regulation of ITMPs.
Katz -
We do have a competitive market in Quebec. Two players - Videotron and Bell - with distinguishable services.
CAIP -
I'm pleased that until today Videotron hasn't applied ITMPs and have increased the capacity of their network. But regardless, there's evidence of duopolistic behavior in our market. This isn't a real distinction between the two. At the retail level, let them decide; just don't force all the wholesalers into a particular model (especially where Videotron isn't…).
Katz -
Well, why don't your customers go to Videotron if Bell is imposing ITMPs?
CAIP -
That's not practical at all; you're not just switching providers, you're switching technologies. There's practical implications of what's sustainable in the marketplace and just switching to the one other provider is not it.
Katz -
In reading GAS tariffs, one provides a dedicated pipe and one doesn't. Yet the tariff doesn't clearly allow for the guaranteed level of service (QOS), or the independence to manage your network as you seek.
CAIP -
Well then I suppose retail ISPs could get a 2-bit service from Bell, but that's not what anybody wants nor what the tariff is for. The guaranteed level of service issue is that no one contemplated ITMPs as the way to ‘guarantee the service'. To say that the tariff reads this way so it is that way is just poor logic.
Commissioner Denton -
The last few days we've been told that the issue is congestion. But you're telling us that it's not the real problem. You said there was no evidence that wholesale was causing a congestion problem; so there's no congestion arising from the offer of wholesale?
CAIP -
We as wholesale customers are not causing congestion - our markets are way too small. If you look at the sequence of Bell applying ITMPs it was first applied to their retail markets, then later to the wholesale. This shows two things - they can segregate the two and we weren't a problem.
Denton -
Drawing on the UK example, you've said that the competitive market will take care of congestion. If retailers were allowed to buy from wholesalers without ITMPs, there would be significantly greater competition?
CAIP -
Yes, the presence of capacity constraint is the effect of duopolistic structure. This is classic business behavior: price as high as possible and constrain. Unified, the infrastructure company in the UK now has the goal of providing service to all companies; the one wholesaler has one wholesale goal. But to increase competition, you need unbundling. It's not just a congestion issue, it's a competitiveness issue. If a wrong policy is implemented on the ITMPs, and later there's unbundling the ITMP policy will have negated the potential of the unbundling to create competition! There's a difference between increased competition and increased differentiation of services.
Denton -
Is publicity at retail level sufficient to handle congestion problem?
CAIP -
Yes. Congestion is caused by users. ITMPs are penalizing behaviours though. What we're trying to address is the behavior; it's extreme to be managing the behavior of all Canadians using a platform when many of them do not have a problematic behaviour.
Commissioner Molnar -
If wholesale services wasn't throttled, could you guarantee that the wholesale users weren't causing congestion to the retailers?
CAIP -
The size of the wholesale market is too small to cause problematic congestion; ITMPs are implemented on a network, not a node or office; so they're implemented as a blanket even though it can be segregated.
Molnar -
You submit that there should be no restrictions on retail and ITMPs are a form of competitive differentiation. You believe that with customer consent, any form of ITMPs should be allowable.
CAIP -
Yes, if there's consent, they can do it. It's a competitive choice for the consumer and for the ISP.
Molnar -
I'm concerned that providing access to applications is becoming confused with competition. I see these as different, the consumer should be choosing.
CAIP -
The only reason packages exist with different features is because there's consumer demand - please don't anticipate a problem that doesn't exist!
Commissioner Lamarre -
I'm concerned with your take on privacy.
CAIP -
There's no need for an a priori rule, there's already laws about it. Is an ISP is going to break the law, PIPEDA or another law will deal with it, and then the ISP may be in double jeopardy before this commission and another.
Chair -
You submit that since we're so small we can't be the problem. And they applied it to themselves and only then you. Doesn't this show you might have been the problem if their initial measure was insufficient? It can be interpreted in both ways.
CAIP -
The problem is exactly that there isn't enough info on this problem, and we shouldn't prioritize one interpretation over another at risk of serious setbacks for competition.
Canadian Internet Policy and Public Interest Clinic on behalf of Campaign for Democratic Media (CDM)
Summary:
- CDM recommends the establishment of normative guidelines and boundaries for ISP behavior, especially for determining whether something is in violation of s.36 or 27(2).
- CDM hopes the CRTC will be forward-thinking, and not reactive, in approaching these matters.
- The problem is the encroaching on the physical and theoretical space of the "public Internet" in service of private concerns and the creation of a false dichotomy between the two.
- When thinking about congestion, both the user and supply side of the equation need to be taken into account in determining solutions. The ISPs are not provisioning to the best of their ability, nor is P2P a reason for the congestion. "Functional marketplaces meet demand with supply - not by squashing demand."
- Traffic interference like application-based throttling is anathema to the open Internet and are unnecessary for the daily operation of the ISP.
- The CRTC must take into account both s.36 and s.7 when determining whether there has been an infringement.
Opening remarks:
David Fewer - Acting Director of CIPPIC
Steve Anderson - Co-founder of the Campaign for Democratic Media and SaveOurNet.ca
Dr. David Reed - Adjunct Professor at MIT
Dr. Andrew Odlyzko - Prof. U Minnesota in Mathematics, leading independent source for tracking Internet patterns
William St. Arnaud - Chief Research Officer for Canada's Advanced Internet Development Organization (CANARIE Inc.)
Anderson
The main issue at stake is who will determine how we use the Internet? The values inherent in the open Internet are values attracting widespread citizen action. If we fail to create the right balance between maintaining networks and keeping the Internet open, there will be grave consequences. The main way to allow ISPs to compete is to increase bandwidth offerings, so they don't monetize on artificially created scarcity.
David Fewer
We are here to convince you of the value of an open Internet. We propose a test for establishing a normative and prospective framework for how to guide ITMPs, and to support competition. This hearing needs to establish a forward looking results, not a reactive tone - we need to set boundaries for ISP behavior with the objective of providing ISPs with competitive security and Canadian consumers and business with confidence in an open Internet. Establishing rules for the open Internet doesn't amount to lawyers running the Internet. No one has a legitimate proprietary claim to the Internet, because it is in the public interest and thus greater than the sum of its parts - it's not the property of the ISPs.
CDM agrees with the objective of this hearing, but takes issue with some of the definitions and assumptions. The CRTC's definition distinguishes between the "public Internet" and "private services" only creates a space for the ISPs to encroach more and more on the public Internet space for the sake of private services. The CRTC assumes that unrestricted traffic increases will lead to congestion and then to the deterioration in services. This simple equation only takes into account the supply side. CDM would replace it with "unrestricted increases in Internet traffic can lead to congestion in all or part of an ISP's network if these increases are not met with adequate provisioning." The ISPs' contention that they are provisioning to capacity but still need to traffic shape all P2P applications is unreasonable in light of the evidence that when Comcast was throttling in order to control growth, it did so by throttling less than 1% of users for no more than 15 minutes at a time. Further, the CRTC assumes that certain ITMP's may be appropriate. The CRTC's language of "traffic management practice" should be referred to as Traffic Interference, because that's what it is at the user end. The language of "integrity of the network" also is imprecise, and doesn't address when an ISP should intervene in the network.
The bottom line is that ISPs have no incentives to tailor their ITMPs to the requirements of the Telecommunications Act.
- P2P specific traffic management is discriminatory against a class of applications and users and is a prima facie violation of s.36 by controlling content and message. This practice also sets up the framework for ISPs to prefer their own services. The purpose of the telecommunication is to speedily pass communication; slowing down P2P applications does just the opposite.
- P2P traffic management is unacceptable. Only peak period traffic amounts to congestion. Targeting P2Pis not in line with the Act. The CRTC cannot be swayed by ISPs' speculation that one day they won't be able to meet demand with provisioning.
- The CRTC must provide a principled framework to guide ISPs in the future. The CRTC should interfere in ISP practices only where it is clear that the market forces can't realize the policy, or ss. 27(2) and 36 are taken into account by an ISP.
Questions:
Chair -
When the ISP sells to the wholesale customer, and there's congestion there, will the provider be affected.
CDM -
It's not a yes or no, it can happen or it might not, it depends on the service the wholesale retailer is giving to the wholesale customer. If the wholesale customers are using more capacity than they had agreed upon, it will degrade their network, and this can have upstream effects. Because the business arrangement might not amount to user capacity, we can run into problems. The best solution would be to have a notification process whereby the wholesaler will approach the customer that they are overusing, and will either increase their capacity or will ask their customers to pay more etc… The issue is solved by business arrangements, not ITMPs.
Chair -
So any application based throttling is unacceptable? I'm not going to make any a priori decisions about what kinds of applications or practices are acceptable or not.
CDM -
Using application throttling is always offside, that's going to be overbroad. There can be many sources of congestion and application-based throttling will have negative downstream effects.
Commissioner Denton -
Some of the criteria developed in the Internet Engineering Task Force (IETF) might be helpful guides here. Is that possible?
Dr. Reed -
The IETF is good at vetting alternatives, providing a broad set of inputs, but its outputs are not regulatory or standards. It has a role to play as a technically neutral forum for discussion but it's not ‘king of the Internet.'
St. Arnaud -
This is really a business regulatory issue, not an IETF issue.
Denton -
Can you weigh in on what we've been hearing about the creation of inadequate capacity caused by insufficient competition?
Dr. Odlyzko -
Demand is not growing very rapidly but is comparable to the rate at which technology is improving. Maintaining a steady rate of capital investment can satisfy increases in capacity to meet demand. There's no need to increase the prices to consumers to meet the increase in capacity.
Denton -
If you had a message to leave us with, what is it?
Dr. Reed - forward looking process that doesn't overreact to what's going on, not overreacting serves the Internet well. Go slow and think forward.
Dr. Odlyzko -
Collect data on a continuous basis, say on traffic levels, and don't go contrary to the findings. At this point, there's no evidence for exponential traffic growth like people are predicting if suddenly all video was watched online suddenly and no other media; if you assess rates of growth for previous technologies too, there's no need for a priori traffic controlling over capacity increasing. I was surprised in my findings to see traffic growth slowing down recently.
St. Arnaud -
The concern is not to fight yesterday's war by targeting P2P. The early adopters are going to be the big users of a new market opportunity. We need to encourage this kind of business development, not slow it down.
Commissioner Katz -
Universities in general consume high bandwidth. What are universities doing to meet their infrastructure demands?
Dr. Odlyzko -
I can't speak about universities as a whole. But at U of Minnesota, there are rules against piracy, and we have attorneys that deal with constant ‘take-down' notices. They have to discuss personally with the users because it's generally been found to be a legitimate use.
Execulink Telecom
Summary:
- There is a need for unique regulation of wholesale different from retail.
- The use of ITMP technologies does not raise privacy concerns.
- The use of ITMPs are generally acceptable for retailers, but disclosure is needed although not necessarily for technical details.
- Applying ITMPs to wholesale Internet traffic is unacceptable.
- It is the duty and right of every ISP to manage their network; allowing competitors to dictate or restrict a competitor's network management is asymmetrical, discriminatory and thus in contravention of the Act.
Opening remarks:
Keith Stevens - Chairman of Execulink
Execulink is a small TSP that offers local phone, cable TV, and Internet access. It purchases some wholesale services (as Bell GAS product) and provides some wholesale services; in both capacities it can bring a holistic understanding of the issues to the table today. Wholesale must be defined because it is different when used as an input for a distinct service or for resale. Wholesale is used to refer to services that are a) resold to customer with no transformation; b) services that are transformed to tailor customers' need, used as an input for a specific service. Wholesale, when used as an input, should be regulated.
Issues:
- Privacy - TSPs have always had access to customer's private info and ITMP technologies don't raise privacy concerns.
- Acceptability of ITMPs- non-technical ITMPs should be the solution of choice. Technical solutions can be helpful. For retail customers, ITMPs are generally acceptable, but with the ISP disclosure to the customer. For wholesalers, it's different.
- Wholesale services - technical solutions when applied to wholesale customers has negative consequences and when combined with the market dominance of DCs, small ISPs are at an undue disadvantage. This is against s. 27(2). These negative consequences have ramifications on the viability of the small ISP businesses.
Questions:
Chair -
Is it possible that wholesale customers cause congestion problems to their retailers?
Stevens -
There's a difference between wholesale and retail that is important here. Wholesale customers buy capacity on volume and speed from the retailer and then the wholesale customer responsibly manages their caps. If the wholesaler is not exceeding these agreements then there is no problem; it's just a contractual agreement both have to uphold. From a retail point of view, if they haven't designed their network to meet their commitments, they have a problem.
Commissioner Molnar -
To what extent are you relying on wholesale facilities rather than your own? Do you use ITMPs on your own network?
Stevens -
It's shifting, now there's more on wholesale but it's shifting to our own. Currently it's roughly 50/50. The mechanisms are there for an attack but they are not being used on our customers yet. We're not even using economic ITMPs like bit caps or anything. Also, we are increasing our capacity just by reinvesting and we've met our demands.
Molnar -
DPI does raise privacy concerns, but your contention is that ITMPs do not raise privacy concerns.
Stevens -
I believe that the current rules and proceedings on privacy are adequate. Privacy concerns on the Internet are no different than telecommunications have dealt with since the telephone.
Molnar -
We've been told by some to prohibit application-based ITMPs.
Stevens -
Some of our customers don't want to download movies, and we could use an ITMP on their package in order to maximize their speed for the services they do want, or for a business person to block gaming etc… But there shouldn't be general restrictions.
Molnar -
You submit that there should be disclosure of ITMPs characteristics but not their technical details. Roks suggested that you need to advertise the throttle speed as well as burstable speed.
Stevens -
With respect to the "up to 25" and are advertising something that is never going to happen, from a network perspective it's very difficult to guarantee speeds. Our customers haven't contracted for volumes. If we did though, there would be a cap.
Molnar -
If there were no ITMPs applied to GAS, could you guarantee that your network users wouldn't compromise others' network integrity?
Stevens -
I can guarantee that we will stay within our contracted volume and speed from Bell. Of course, we all affect each other, but the contracts are there to govern it.
Molnar -
Is there anything particular to the fixed wireless technology that we should know with regards to ITMPs?
Stevens -
Fixed wireless providers have an ability to affect each other more, so you might want to put ITMPs on it before another type. The difference is in the timing of the need for ITMPs.
Commissioner Denton -
You contend that your reinvestment (appreciation rate) is enough to cover the cost of increasing infrastructure capacity, at what rate is it?
Stevens -
The appreciation rates are also relative to how long we think the equipment will last. But yes, it is enough.
Primus Telecommunications Canada Inc.
Summary:
- All ISPs should be able to manage their networks and employ ITMPs as they see fit.
- All ITMPs are generally acceptable and with consent, there is nothing a priori wrong with them.
- If users do not change their habits though, the commercial ITMPs are useless
- Primus does not use aggregated information; DPI is only for network management and planning in congested moments
Opening remarks:
Matt Stein - VP of Network Services
Operates as a reseller but makes significant inroads in innovation. They were the first to employ VOIP. They employ a DPI two-part process. First, determine the nature of the traffic to prioritize it and second, classify and prioritize it on the network. When there is no congestion, there is no prioritization even if only for a few minutes. They don't employ application-based throttling, but only during the moments of congestion. Primus is committed to increasing capacity, but ITMPs are necessary and complementary with a holistic approach to network health. ITMPS are only for unexpected surges in traffic. Primus is not a broadcaster, does not produce content nor is it affiliated with content creators. DPI ITMP is only for network management and not for aggregate information storing, except information on network usage for network planning purposes.
All ISPs should be able to manage their networks and employ ITMPs as they see fit. All ITMPs are generally acceptable and with consent, there's nothing fundamentally wrong about them. If users do not change their habits though, the commercial ITMPs are useless and will lonely result in increased costs. Our customer feedback tells us that our DPI usage is utilized well and in a balanced manner so that their service is as they want.
Primus believes that all ISPs should disclose their ITMPs on both retail and wholesale services, but not about the security related ITMPs. Upstream retailers should be required to notify downstream customers of the ITMPs and detail the foreseen impact on the downstream customer's service. Primus discloses DPI usage in advance, on its Web site. But when Bell used the ITMP without disclosure to Primus, it had a negative impact on Primus' customers and resulted in much dissatisfaction.
Upstream ISPs should not be permitted to impose traffic management practice upon their downstream ISPs as it hinders innovation and precludes the ISP from giving adequate service across regions and customers.
Questions:
Chair -
You don't share the view that the application-based control should never be allowed?
Primus -
It's between an ISP and its end-users. End-users need to have choices, and ITMPs are part of this informed choice. If an application were very highly valued on the network, they should receive priority whatever it is.
Chair - you're a wholesaler provider and customer. As a customer from Bell, should they be able to use ITMP?
Primus -
No, an ITMP should only be allowed between an ISP and its direct customer, an ISP and the end-user. Our direct customers are not customers of Bell it is unacceptable that Bell should affect our direct customers.
Commissioner Katz -
Have you asked Bell for the type of notification you recommend?
Primus -
Yes we have, but we have no answer.
Katz -
So you only apply your own QOS on your own infrastructure?
Primus -
Yes, a customer can pay the same amount of money for the Bell DSL service and the Primus DSL service, and our customers can see a difference between the two, and thus transferred them to the Primus ONNET. The nature of the QOS is how we apply ITMP.
Commissioner Molnar -
As a content provider, you're offering two different experiences for someone streaming and someone downloading a P2P file.
Primus -
People have different experiences of video - streaming begins within a few seconds, but downloading P2P it doesn't begin immediately. Yes, there is a different caliber being offered for these two applications, but the user expects that to be different based on how they work, not because of the ITMP.
Molnar -
An ISP should deliver content, and not compete on what content is available and when moves you from being a communications carrier to a "semi-dumb pipe."
Primus -
We are still building capacity and there isn't a blanket ITMP. We only use the ITMP in a discrete instance and place.
Molnar -
You say you disclose your ITMP. So if I go, what does it explain to each customer? And is it in a way that a customer can understand?
Primus -
We'll include a copy of what's online in our final filing; it was written for the customer to understand and explain why it's good.
Commissioner Lamarre -
You explained that the most popular applications get high priority. This is inevitably going to disadvantage a customer using less popular applications, marginalizing them? And most popular applications begin as marginal ones, don't you see it as counterproductive then?
Primus -
Well, at most times of the day, all applications are flowing just fine; we only use ITMP in discrete instances and not as a blanket on applications. Indeed, new protocols maybe only given ‘normal priority' at moments of congestion, but they will become popular as they are used during other moments when there is no congestion.
Lamarre -
For the customer who disagrees with the implications of the DPI, is there another choice of package or do they have to change carriers?
Primus -
No, there was no other offering.
Lamarre -
Privacy being a personal right, we need to be concerned about this. You say you don't collect the information except for network planning. But the fact remains that with DPI you're "snooping in." Where is the assurance that you did anything else prior to invading privacy?
Primus -
Because of the way we apply it, this is the least invasive on customers' usage of the Internet, which would lead to great dissatisfaction. Because we don't collect information except in aggregate form, we don't feel it's an invasion of privacy.
Molnar -
How difficult is it to change your policy rules on DPI?
Primus -
We spend more time making sure that our rules won't upset our users or affect their usage. If you've deployed it in a varied distribution, rigidly, then you might need to make some changes to the network physically or otherwise. It's all to ensure the customer satisfaction. The overwhelming majority didn't notice that the network was congested at all during the Michael Jackson memorial because of the targeted way our ITMP is deployed.
Translate
