Home » Internet

Improving on Amazon’s Net Neutrality proposal

By 19 January 2010 2 Comments

Amazon filed a response to the FCC’s proposal for Net Neutrality rules and the response has raised some eyebrows given the fact that Amazon is one of the signatories to the Open Internet Coalition letter in support of new regulations on the Internet.  In the context of Google’s softening on Net Neutrality which now recognizes the right of network operators (and not just Internet users and sites), this may actually be the beginning of an era of meaningful dialog that could eventually lead to some sensible but not overbearing regulations.

Right off the bat, Amazon acknowledges the fact that content providers have long been able to buy better access through the use of things like edge caches.  This is already a radical departure from the hard line Net Neutrality advocates like Free Press that insist on the delusion that “all websites and applications download and upload at the same speeds”.  But the most significant aspect of Amazon’s response is the stunning acknowledgement that ISPs should have the right to sell end users OR Content, Application, and Service (CAS) providers as well as the devices they offer enhanced performance.  Digital Society has also sent a reply to the FCC which makes a similar case for allowing ISPs to offer premium services.

Amazon even goes as far as proposing some modifications to the language in section 8.13 of the FCC’s proposed Net Neutrality regulation.  The following excerpt of Amazon’s letter to the FCC shows the changes (underlined text) that Amazon would like to see the FCC adopt.

8.13 Nondiscrimination and No Harm

Except as described herein, and subject to reasonable network management, a provider of broadband Internet access service must treat lawful content, applications, and services, and devices in a nondiscriminatory manner. A provider of broadband Internet access service may, for compensation on nonexclusive terms:

(a) Offer enhanced quality, speed, or other functionality to individual providers of lawful content, applications, services, or devices, so long as doing so does not degrade the quality, speed, or other functionality provided for any other lawful content, applications, services, or devices, from any source or for any user; and

(b) Offer individual users the option of enhanced quality, speed, or other functionality for specified lawful content, applications, services, or devices, or any source thereof, provided that such enhancements shall not degrade the quality, speed, or other functionality provided to other users.

The fundamental argument that Amazon makes is the “no harm” argument.  Amazon argues that because performance enhancing edge caches don’t harm uncached content, then it is an acceptable arrangement.  And if ISPs want to offer other forms of performance enhancements to end users or CAS providers, then it should be permissible so long as the services are on “nonexclusive terms” and they do “no harm” to other CAS providers or end users.  Granted, the CAS providers or end users who don’t pay for these performance enhancing services won’t get any kind of performance boost.  They simply remain in the lower performing state they were in to begin with.

The problem with the “no harm” definition

Amazon’s proposed revisions are certainly a welcome step in the right direction, but there are significant problems that remain.  The definition of “no harm” needs to be qualified or else it becomes an unobtainable requirement.  That’s because any time we add an additional CAS provider service to the network or we add any additional broadband subscribers, there is always going to be “harm” to other CAS providers and other subscribers.  That’s because in any finite capacity network, any additional load on that network will cause it to have less available resources to the existing user base which by definition means it is degrading the network (at least until the network operator adds more capacity).

Note that in the real world when enough users and CAS providers start buying services and a certain threshold for congestion is reached, the network operator must increase capacity to give their customers the level of performance necessary to keep them and attract more customers.  Furthermore, the fact that there are more paying subscribers or CAS providers will mean more available revenue for fund more infrastructure.  This is consistent with the fact that Internet services have continuously improved over the years.

The problem with user rights

The first thing that needs to be fixed is the concept of “user” rights.  To really be fair, we need to be talking about “subscriber” rights.  That’s because some subscribers might have a family with 6 Internet users sharing the same connection and another subscriber might only be a single Internet user.  Does that mean the home with 6 users should get 6 times more bandwidth than the subscriber with a single user if both subscribers are paying the same?  Certainly not!  Subscribers should get a share of capacity proportional to what they pay.  A broadband subscriber who pays twice the norm should get twice the minimum performance (during peak congestion times) and possibly more than twice the burst performance due to volume discounts.  We all accept the fact that bulk purchases cost less per unit than small quantity purchases and we see this in every facet of life and in every super market and the Internet is no different in its usage of second degree price discrimination.

This concept of subscriber rights versus user rights is even more relevant to peer-to-peer (P2P) usage.  If subscriber A is on average communicating with 100 other P2P users on the Internet, and subscriber B only talks to one server on the Internet on average, does that mean Subscriber A deserves 100 times more bandwidth than Subscriber B even though both subscribers pay the same amount for broadband service?  Most reasonable people would say that all subscribers deserve the same share of bandwidth, and it might even sound redundant to them that I would bring this up.  But some of the more hard line P2P users actually believe that each of the users that they’re communicating with deserve an equal share of the broadband network’s capacity which justifies their taking of 100 times more bandwidth than other broadband subscribers.  That’s silly of course because the broadband provider’s responsibility is to its subscribers who pay for the network, and not to the hundreds of users that some individual subscriber might be communicating with.

A fairer and more detailed proposal

So the better definition of “no harm” would be “no disproportionate harm”.  If a new application or broadband subscriber starts using the network, they should be granted a share of the network in proportion to what they are paying for.  To get really technical, the amount of capacity “C” allocated to an application or subscriber should be:

C = X times Y divided by n where

  • C = Capacity allocated (capped by contracted bandwidth)
  • X = Total capacity of network
  • Y = Tier coefficient determined by how much a subscriber and/or content provider pays
  • n = number of active subscribers or applications

The formula is used a bit differently depending on whether it is being applied to inter-subscriber capacity allocation (allocation of resources between subscribers) or intra-subscriber capacity allocation (allocation of resources between users of a single subscriber).

In the case of inter-subscriber capacity allocation, “Y” might be solely determined by what the subscriber is directly paying the broadband provider or it might also include a subscriber’s indirect payment to the broadband provider in the form of payments from the content provider.  In the latter case, the value of “Y” would increase due to the direct and indirect revenue coming from the subscriber.  So if the Y value is greater, then the share of capacity allocated should also be greater.

In the case of intra-subscriber capacity allocation, “n” is just determined by the number of active applications and “Y” doesn’t need to be considered.  While intra-subscriber network management can actually be performed to some degree by the broadband subscriber to be neutral or non-neutral depending on user preference, the reality is that most subscribers don’t have the capability or the will to learn such complex technologies.  In light of this reality, there is nothing wrong with broadband providers offering neutral intra-subscriber network management with or without additional charges.

So if a broadband provider wants to manage their network (with or without additional fee) such that VoIP or gaming applications always get priority over everything else, that’s perfectly reasonable because it fits the concept of true neutrality.  That’s because VoIP and gaming never come close to using their fair share of the network because they’re low bandwidth and typically consume less than 0.1 Megabits per second (Mbps).  How can it be unreasonable to protect these jitter-sensitive applications by carving out 0.1 Mbps and leaving 1 to 20 Mbps for every other application?

Even when applications like VoIP and online gaming are given the highest level of packet forwarding priority, they don’t deprive other applications of any noticeable amount of bandwidth and they don’t induce any noticeable latency or jitter on other applications.  On the other hand, P2P applications like BitTorrent can deprive other applications of their fair share of the network and cause massive jitter problems even though BitTorrent claims to have a network friendly protocol.  The problem now is that the proposed FCC regulations might ban this type of good network management because it’s getting some bad technical advice about network prioritization technology from groups like Free Press.

User approved prioritization and discrimination

But what if a content provider like Netflix wants to guarantee that its video on demand applications should always be guaranteed up to 4 Mbps on a broadband subscriber’s connection regardless of how many other applications are running on the service?  That means if 3 other applications are running and the total capacity of the network is 6 Mbps, those three other applications will have to share 2 Mbps while Netflix gets a guaranteed 4 Mbps.  On the face of it, that might sound like a bad form of discrimination.  But what if the user wants this kind of behavior so that they can have a trouble free viewing experience for content that they are actually willing to pay for?  What if the user reads a clear notice from Netflix asking them if they want this kind of favoritism for Netflix on their own broadband service and the user opts into the service?  Shouldn’t the user have the right to favor some applications and content over others for their own viewing pleasures?  Of course they should!

The current reality is that caching seems to be good enough for most buffered web video content so long as the subscriber doesn’t use other network intensive applications while watching the video.  But the future might change as more homes have more simultaneous active devices using their broadband service.  Why foreclose the possibility of this type of service and/or network management especially when the broadband subscriber is asking for it?


While we don’t completely agree with Amazon’s proposal, we believe it’s a great first step to better dialog and we can finally set aside the bumper sticker slogans and get to the real debate.  Amazon has taken a huge step in the right direction but their definition of “no harm” and “user” needs to be clarified to account for this issues brought up in this article.