Discussion:Some believe that at the core of the NN debate is the ability to price discriminate and to offer different tiers of service. [Odlyzko Efficiency and Fairness 2009 45] The issue of bandwidth caps had its origins in the late 1990s when AOL when to a ~$21 all-you-can-eat dial-up offering. The competitive pressure from this offering, tied with AOL's carpet bombing of the market with free AOL installation CDs, forced other ISPs to migrate away from metered bandwidth consumption to all-you-can-eat offerings too. While these offerings were "all-you-can-eat," they were not unlimited. The subscriber could only use the service up to the bandwidth offering of the telecommunications. Dial up ISPs in the 90s were making the slow migration from 14.4 kbps modems to 56 kbps. At those rates, it was hard for the subscriber to do significant damage with hogging bandwidth (subscribers did tie up phone lines which led to an access charge debate, but this is another matter).
The broadband era began slowly in the late 90s with minimal broadband offerings. First cable modem service was offered and then DSL service. But both of these initial services were of limited broadband capacity. Power users could certainly use more broadband, but the disparity between light users and heavy users was not yet as pronounced.
Eventually DSL and Cable increased the capacity of their lines, increasing the ability of subscribers to use the service. Now the difference between a power user and Auntie Em checking email became more pronounced. Further, as the cable infrastructure is a shared infrastructure (all users on a cable network share the bandwidth from their home to the cable headend - where all DSL and fiber lines are direct connections from the home to the core network and are not shared), any users which consumed a disproportionate amount of bandwidth on the cable network congest the network and impact all other users of that cable network.
Finally, a few providers such as Verizon began to offer Fiber to the Home (FTTH). These are significantly higher capacity offerings where the difference between a power users and a casual user is pronounced.
Nevertheless, in all of these offerings, from dial-up to fiber, even though the offering is "all-you-can-eat," the offering has always been limited or capped by the capacity of the telecommunications infrastructure.
The issue with tiers or "broadband caps" as they are called is that they are artificial caps set artificially lower than the capacity of the telecommunications infrastructure being deployed.
Previously, some cable networks terminated customers who exceeded bandwidth limits, without reported informing customers what those limits were. Recently, some carriers have begun to impose bandwidth caps on users - individuals who exceed the bandwidth cap will be forced to pay additional fees.
- Comcast has announced a 250 Gbpm cap on bandwidth consumption. [Bangeman 091707 (stating that Comcast is seeking to move heavy bandwidth users to 'commercial grade' services)] [Borland 092203 (Comcast sent letters to heavy users stating that they were using too much bandwidth, without providing notice of what "too much bandwidth" means, and stating that the users might have their service cut off)] [Enoch ("The company has a bandwidth limitation that, if broken, can result in a 12-month suspension of service. The problem, according to customer complaints, is that the telecom giant refuses to reveal how much downloading is too much.")
- Time Warner Cable [Bangeman 011808 (reporting that TWC would be initiating "Consumption Based Billing," and noting that Comcast had cut service off to subscribers that had reached bandwidth limits, but that Comcast did not provide notice of what that limit was)]
- Some wireless services are announcing caps on their services.
"One percent of the company's customers account for 20 percent of the network usage; the top five percent account for 40 percent of the usage. Because the network must be able to accommodate peak traffic loads, AT&T -- like other network providers -- finds itself building far more capacity than most users need, just to support the most prolific users." - John Donovan, CTO of AT&T, in Wired.com |
History: Internet assumes shared resource. But at times demand exceeds supply. Congrestion control has traditionally been done by TCP. Technologists did not look to economic solutions; technologists look to technological solutions to congestion. Over time congestion control has evolved to new tools:
- Volume Caps that limit the total volume of traffic over different durations of time
- Prioritizizing subscriber or application traffic
- Rate limiting traffic classes
New ideas in tech community
- Re-ECN
- P4P, IETF ALTO Working Group
-
Advocates of Tiers point to service
level issues and make "tragedy of
the commons" type arguments. They suggest that Network Neutrality restrictions would mandate that carriers would be obligated
to charge the same amount for all individuals: lite users as well as
bandwidth hogs. The carriers argue that a small number of users, specifically P2P users, are responsible for 90%~ of their bandwidth usage. They also point to YouTube and Peer-to-Peer users as causing network congestion. [FTC Staff Report 2007 p 62] Differentiating bandwidth caps for different users, they argue, means that they could distinguish between the bandwidth hogs consumer the majority of resources, from the average user consuming a minority of resources. [Farber] [Pepper] If networks cannot discriminate, bandwidth hogs will use up all available bandwidth, cause a burdensome cost on the network service provider, and thereby eliminate the incentive of the networks to build. [Sidak p 5]
Adam Thierer at PFF advocates for what he calls "Ramsey Two-Part Tariff" where services is offered for a flat rate up to a specified cap, and then after that cap service would be on a per-unit basis. [Thierer]
Some NN Proponents note that the offering of different levels of service
does not violate traditional common carriage jurisprudence. [Odlyzko Efficiency and Fairness 2009 48 (noting senior citizen and student discounts, stating 'what was illegal was "undue or unreasonable" preferences')] The classic
case in this area involves a telegram that was transmitted with a
typographical error that resulted in harm to the customer's business.
The customer sued. Western Union in its defense noted that it offered
different levels of service and insurance. In order to ensure against
error, for an additional fee the message could be repeated one time or
even two times, with corresponding levels of insurance for each level.
But Western Union disavowed liability when no error correcting service
level was purchased. The customer complained that this violated common carriage principles. The court held that as long as Western Union offered a service level where it was responsible for the proper
delivery of the message, it was free to offer a discounted service
level where the message was not guaranteed to be error free. [See Liability;
Union
Pacific v. Burke] In sum, the precedent is that as long as
there is a service that comports with the principals of common carriage, variations of that service offering with financial benefit
are permissible. (See also different levels of service offered by
the common carriers the US Post Office and Airlines).
Internet access service has always been tiered, where the tiering was tied to the physical transport medium.
- Dial-up service restricted to a maximum of 56 kbps was sold at one rate;
- DSL service at different maximum bandwidths are sold at different rates;
- cable service are sold at different rates; and
- Higher bandwidth T1 or fiber services are sold at a different rate.
No one has suggested that DSL service must be sold for the same price as dial-up service, even where the same copper wire might be used. They key, however, is that the upper limits of all of these services had been limited by the nature of the physical medium; you can only get so much bandwidth out of a dial-up line.
A new problem arises with pipes where the physical transport medium limits are high. Should the service provider be able to provide lower bandwidth for one price and higher bandwidth for another. What about high bandwidth, high consumption (bandwidth hogs?). It is said that some proponents object to this type of tiering. [Weiser ALR 08 p 4] Apparently Senator Wyden introduced a legislative proposal that would have banned tiering of this type. There is little, however, in the history of common carrier or of Internet access that would support this objection.
David Clark
notes that the cost of broadband is cents for the GByte ((a) under 10 cents per GB for transit, (b) 5 cents per GB for direct connection to content providers, (c) 2 cents per GB for local traffic), and that if one accepts the assumption that heavy users are subsidizing light users, the subsidy is in the form of pocket change. Clark argues that access networks can bring down costs by avoiding (a) transit costs through (b) direct negotiations and connections to content providers, or through (c) local caching. [Clark TPRC 2008] The cost is not zero, but it is a small number (rural ISPs like WISPs, its more expensive).
Some NN Proponents of network neutrality argue that this is a false scare tactic. As noted above, different service level agreements is the norm. If bandwidth hogs require more bandwidth, charging more for the service is not a problem. Traditionally, different levels of bandwidth have been offered at different rates. Internet access at dial up speed was sold at one rate; DSL speed was sold at another rate; and a T1 Internet connection was sold at another rate. Proponents argue that this is not an argument that everyone should pay the same rate; this is an argument that there should be some basic neutral Internet connectivity over which individuals can interact with other servers, utilize any application, or interact with any content - more than that, service providers can offer specialized offerings. [Annenberg #3] [Lessig Testimony 2008 p 4 "While broadband providers should be free, in my view, to price consumer access to the Internet differently - setting a higher price, for example, for faster or greater access - they should not be free to apply discriminatory surcharges to those who make content or applications available on the Internet."]
A reaction by networks to bandwidth hogs has been to attempt to block something associated with the hogs. As noted above, P2P networks have been identified as part of the bandwidth hog problem, and some networks have blocked P2P software. Proponents note that blocking P2P is a blunt tool; it blocks legitimate traffic along with the complained of traffic.
Proponents argue that the issue is how to do this in a fair way that promotes innovation and encourages, not discourages, broadband usage. Greater bandwidth will lead to greater innovations including the potential of new video opportunities. Inappropriate caps and discriminatory pricing will discourage bandwidth consumption in a way that could inhibit the next generation of applications.
Some NN Proponents object to tiers and express concern that it will limit innovation opportunities. [Malik] One of the problems with bandwidth caps is that - while they may be caps only excessive users could reach today - at the rate of growth of broadband use and as average joe hockey mom's increasingly incorporate the Internet into their lives - they are caps that the average joe hockey mom will slam up against in the near future. [Malik]
Peak / Non- Peak Usage
Noah Pepper makes the argument that bandwidth caps are too clumsy to be appropriate. He argues
Excessive bandwidth usage is only a problem when it degrades the quality of service for other users of the network. Service is only likely to be degraded for other users during peak usage hours because that is when the most people are on the network . . .
Targeting these alleged bandwidth hogs for having large total usage by restricting network access or raising rates doesn't get at the real problem, which is congestion during times of heavy usage. These penalties are essentially punishing consumers for gaining extra utility from the service, particularly at non-peak usage hours, at no harm to the ISP or other users . . .
Throttling everyone during peak usage is the only sensible solution to this problem (besides upgrading the network).
[Noah Pepper] [See Lee (concurring with off-peak pricing)] In other words - identify the problem - the problem is congestion on the network during peak hours with perhaps a few users place unproportional demand on the network - the problem is not usage of the network during not peak hours when the network is not congested. The solution to the problem - just like the mobile phone service providers - is to provide a scheme that charges for peak usage and creates an incentive to migrate traffic to non peak hours.
Not Network Neutrality:
In time it appears that the issue of tiers and bandwidth caps is being separated out from the Network Neutrality debate. While similar people may or may not like tiers, Network Neutrality is fundamentally an issue of discrimination (and common carriage). Bandwidth caps and tiers do not necessarily offend anti-discriminatory notions.
