Save the Internet Act Doesn’t

Let’s face it, the net neutrality debate has been bad for the public’s estimation of the American process of technology policy creation and enforcement. This issue is a twenty-year-old sideshow, played out in public with great fanfare while the Internet quietly developed into a very dangerous and scary system.

Originally intended to regulate the Internet according to its unique, carrier-neutral, end-to-end managed design, net neutrality has come to mean the opposite: the current proposal by Congressional Democrats uses telephone law to force Internet access into a model most familiar to telephone regulators. This policy shift benefits experts in telephone law, but it does little to protect consumers from abuse.

It’s also out of sync with voter concerns. While a widely publicized push poll purports to show massive consumer support for net neutrality (defined a an idiosyncratic way), net neutrality ranks far below the key issues (such as privacy and robocalls) voluntarily listed by consumers as their actual concerns.

The Robocall Irony

Robocalling is hugely annoying to consumers. It threatens the viability of the entire telephone network by polluting authenticate calls with spam. Robocalling is enabled by an open Internet in which certain users do what they want without interference by carriers.

Consumers want carriers to stop robocalls and carriers are willing. But progress on this issue depends on systematically violating two core principles of net neutrality: carrier indifference to the sources and content of network communications.

Robocall prevention depends on an emerging technical standard known as the STIR/SHAKEN framework that authenticates caller ID information for VoIP phone calls. Authentication at network interconnection points should have been part of the original design of the Session Information Protocol (SIP) used by VoIP. But it was too hard, so the problem was papered-over with an inadequate end-to-end design and deferred.

Blocking Malicious Callers Isn’t an End-to-End Problem

Authentication should also have been at least an optional portion of the design of the Internet Protocol as well, but it wasn’t because that would have made the system too complex for the hardware on which the Internet was originally deployed in the 1970s.

The original Internet was also a closed system governed by a strict acceptable use policy, so criminality was prevented by social firewalls. You simply weren’t allowed in unless you were known to be trustworthy to the central authority that administered the network, the National Science Foundation.

Opening up the Internet to the general public took the central authority out of the picture and also welcomed criminals onto the Internet. The Internet community has struggled with spam, identity theft, and extortion ever since, always with limited success.

We’re never going to have another NSF deciding who can and can’t get on the network, so we’ve devised piece-by-piece replacements for its administrative functions through mechanisms such as the SIP Forum and gatekeeper businesses such as Cloudflare that regulate interconnection.

Net Neutrality is a Naive Policy

Advocates of net neutrality often claim that it was part of the Internet’s original design. This is true to the extent that we overlook the roles played by the administrators of ARPANET and NSFNET, the organizations that established the standards and practices for transmission, routing, and interconnection.

Accepting this claim doesn’t suggest that the Internet of the present and future has to follow the original mold. A fully inclusive Internet faces new problems that demand new solutions driven by new users and new uses.

Net neutrality presumes that network interconnection is a simple problem amenable to traditional solutions. While I wish this were the case, it hasn’t been true since the decommissioning of NSFNET in 1995.

Paid Prioritization is Both Good and Bad

The Internet’s technical standards for network interconnection are complex and evolving, but the Obama FCC’s Open Internet Order of 2015 treated interconnection as if it were a common carrier problem. While stopping short of enacting detailed rules, it classified interconnection under Title II and reserved the authority to police it as it saw fit [para 193]:

As a result, the Commission will be available to hear disputes regarding arrangements for the exchange of traffic with a broadband Internet access provider raised under sections 201 and 202 on a case-by-case basis: an appropriate vehicle for enforcement where disputes are primarily over commercial terms and that involve some very large corporations, including companies like transit providers and CDNs, that act on behalf of smaller edge providers. However, for reasons discussed more fully below,[fn 479] we exclude this portion of broadband Internet access service—interconnection with a broadband Internet access service provider’s network—from application of our open Internet rules. We note that this exclusion also extends to interconnection with CDNs.[fn 480]

Hence, interconnection agreements and practices with the effect of “prioritizing” some companies and applications over others – nearly all of them – are unreliable under the proposed bill that the House will vote on tomorrow. This is to be done in the name of creating a level playing field, a deceptive claim.

The Internet is Tilted by Design

The Internet of the NSFNET era was essentially level because it was only one network with multiple users, each of whom had the same form of access. Today’s Internet is a loose confederation of multiple networks of varying degrees of performance.

Hypothetically, Google and Facebook may choose to over-provision their networks so that their servers are rarely over-utilized.  Netflix and Amazon may chose not to because it would cost too much, so they would deal with overload by temporarily reducing video quality.

Neither approach is right or wrong, and both are commercially reasonable tradeoffs reflecting company strategies about investment, growth, and free cash management. The dynamic and evolving character of today’s Internet means that goals built on historical designs are no longer beneficial.

Limits of Fairness

Net neutrality is an appealing idea because it seems to be about fairness: as long as every interconnection, every packet, and every application is treated just like all the others, the Internet must be fair. This may be fair, but it’s neither just nor efficient.

Firms that build their own network infrastructure – such as the Tech Big Five in market cap – can prioritize their traffic as they see fit all the way from their servers to their ISP interconnection points. They do have to contend with traffic from other companies over the ISP last mile, but last mile segments (especially on cable and fiber networks) are over-provisioned.

Firms that depend on shared CDNs and transit networks are much more susceptible to variations in quality because they’re sharing with each other from their content’s point of origin all the way to the consumer. So the best we can do to for fairness on today’s Internet is to ensure that firms with similar resources and network scale are treated similarly.

Congress is Trying to be Too Clever

The US is stuck in a chicken-and-egg mode with Internet regulation. Congress and the regulatory agencies are supposed to have defined roles, where Congress sets the policies and the agencies do the enforcement. When the FCC creates regulations out of whole cloth and Congress then puts these inventions into law verbatim, we’re not doing things the right way.

The Save the Internet Act of 2019 takes this backwards approach a step further by codifying the “general conduct standard” from the 2015 Open Internet Order. This controversial provision either puts “a cop on the beat” and/or massively violates the Administrative Procedure Act depending on where you sit.

I find it problematic because the rule-making process in the APA is supposed to ensure that firms know what the rules are. A rule that says the regulator can sanction firms for conduct that doesn’t violate any known rules runs counter to the expectation of certainty the APA creates. This is like Congress saying “we will punish you for doing things we don’t like but we’re not going to tell you what those things are,” a rudely passive-aggressive recipe for confusion.

The Conflict in the Open Internet Regulations

The general conduct standard is short but vague, with a lot riding on the regulator’s idea of reasonableness (§ 8.11):

Any person engaged in the provision of broadband Internet access service, insofar as such person is so engaged, shall not unreasonably interfere with or unreasonably disadvantage (i) end users’ ability to select, access, and use broadband Internet access service or the lawful Internet content, applications, services, or devices of their choice, or (ii) edge providers’ ability to make lawful content, applications, services, or devices available to end users. Reasonable network management shall not be considered a violation of this rule.

Right off the bat, it’s in tension with the ban on paid prioritization because network management needs to differentiate traffic (§ 8.9):

(a) A person engaged in the provision of broadband Internet access service, insofar as such person is so engaged, shall not engage in paid prioritization.

(b) “Paid prioritization” refers to the management of a broadband provider’s network to directly or indirectly favor some traffic over other traffic, including through use of techniques such as traffic shaping, prioritization, resource reservation, or other forms of preferential traffic management, either (a) in exchange for consideration (monetary or otherwise) from a third party, or (b) to benefit an affiliated entity.

Now let’s see how these regulations interact with each other.

The Conflict Scenario

Here’s an example of a conflict scenario complicated by the regulations:

  1. A VoIP provider sends a stream of packets to a transit provider that are evenly spaced.
  2. The transit provide combines, on a higher speed data connection, this stream with traffic from a video stream that consists of evenly spaced clumps of packets.
  3. The video clumps distort the spacing of the VoIP packets.
  4. The ISP forwards this combination of packets to a lower-speed neighborhood network.

The ISP has a choice about how it does its part of the job; it can either forward the packets first-come, first served as the net neutrality advocates want, or it can reorder them so that the VoIP packets are forwarded according to their original timing, as they were before the transit network combined them with the video clumps.

Preserving the original timing is preferable because it’s good for call quality, but it’s not default behavior where a transit network is involved. The big private networks (Google and Facebook) can do it, and so can ISPs.

Resolving the Conflict in Favor of  Consumers

In either case, costly additional processing is required, and in many cases the call will be fine with default behavior. It’s harder for the ISP to detect every case that will benefit from non-default treatment than for the big private network because the private networks tag traffic according to transmission requirements.

ISPs may be allowed to take steps to improve third party call quality as an exercise in reasonable network management, depending on how were interpret the two contradictory regulations just cited. If VoIP providers were allowed to contract with ISPs to identify and optimize their traffic, it would be a win for many: consumers would have more choice and better quality, and ISPs would have more revenue.

Google and Facebook would suffer because the network investments they’ve made wouldn’t be as valuable as they are today if the little guys could pay a few cents per call for better quality. Guess who’s in favor of the ban on paid prioritization?

The 2015 Open Internet Order Must Not Become Law

The Save the Internet Act‘s authors say a vote for their bill is a vote for net neutrality because the regulations it would freeze into law are so excellent. But the reality is that the 2015 OIO is terribly confused because its drafters don’t understand how the Internet works.

As Carl Sagan said in his last interview: “[W]e’ve arranged a society based on science and technology in which nobody understands anything about science and technology, and this combustible mixture of ignorance and power, sooner or later, is going to blow up in our faces.”

It’s reasonably clear that Internet regulation is now blowing up in our faces: Congressional Democrats are intent on raising the 2015 OIO from the dead, but for reasons that appear to be totally political. Meanwhile, data brokers make hay with our browsing histories and nobody but the Europeans seems to care.

Perhaps the time has come for some sort of symbolic resolution that puts net neutrality to bed. Congress needs to get on to things it understands.