Throttle-gate: What are the Facts?
A lively discussion broke out on the Internet last Friday over the Wall Street Journal’s revelation that Netflix has been secretly throttling video streams sent to customers of mobile broadband services but not to any wireline or satellite carriers. Netflix was forced to issue an explanation on its company blog, and a small number of Netflix allies and media fans stood up for the company and the practice of throttling. Foes of the FCC’s Title II regulations were less generous and understanding.
The issue covers three aspects of the regulations that currently cover Internet Service Providers in the US today as a consequence of last year’s Open Internet Order (OIO) passed by the FCC’s Democratic majority:
- The OIO’s hard line rule against unreasonable throttling;
- The scope of the OIO with respect to the kinds of service providers covered and exempted;
- The OIO’s transparency rule.
This post is an examination of the factual elements of the controversy, particular with respect to the engineering logic of the practices that Netflix has admitted to doing. I take it for granted that Netflix has been throttling at least some mobile streams because their blog post admits as much: “…our default bitrate for viewing over mobile networks has been capped globally at 600 kilobits per second.” The Wall Street Journal finds that AT&T and Verizon are throttled more than others, however, so Netflix does not appear to have come clean.*
So why would Netflix throttle mobile broadband speeds? As it turns out, there is one and only one good reason to do this on an a priori basis: to save Netflix’s mobile customers money. Throttling data rates in this way reduces the volume of data sent to the consumers, which helps keep consumers under their monthly data limits and provides them with a bit of insurance against overage charges.
This isn’t the only form of throttling Netflix does, actually. The firm – like every other video streamer on the Internet – also throttles its data rates when congestion limits its ability to consistently stream at a higher rate. But that sort of “dynamic rate adaptation” is very different from imposing a hard limit on video streaming at such a low level. The 600 Kbps ceiling is barely higher than the absolute minimum rate at which Netflix will operate, 500 Kbps.
While dynamic throttling aims to deliver the highest picture quality consistent with avoiding buffering delays, a priori throttling ensures that the picture quality will never be very good. For many – perhaps most – Netflix users watching video on smartphones while in transit, 600 Kbps might be good enough, but for those who use newer phones with very high resolution screens, it’s noticeably clunky.
People who have low data plans and who want to watch lots of videos benefit from the 600 Kbps limit, but those with more generous plans and those who watch very little video outside of Wi-Fi networks have actually been harmed by them. The Netflix system is a one-size-fits-all solution to a problem that doesn’t fall equally on all of its users.
Strangely, the throttling also has the peculiar side-effect of reducing data revenues to mobile carriers. No one is complaining about that, but throttling carrier revenues does have a limiting effect on infrastructure build-out. But that’s a side issue. Preemptive throttling doesn’t do anything intelligent for congestion, but adaptive throttling does.
Who is in the Best Position to Throttle?
Strictly as a technical matter, throttling can be done by at least three different parties every time a video steam is communicated from Netflix to a user: in addition to Netflix itself, streams can be throttled by transit providers or by the mobile networks themselves.
At the time of the infamous John Oliver rant on net neutrality, one of the Netflix transit providers, Cogent, was throttling Netflix streams to wireline ISPs such as Comcast.
While Oliver and Netflix (and Netflix lobbyist Marvin Ammori) blamed the throttling on the ISPs, Dan Rayburn determined that Cogent did in fact selectively de-prioritize high volume streams when it found itself unable to deliver service it had sold to Netflix. Cogent created fast lanes and slow lanes and Netflix blamed the whole thing on ISPs.
While there are good reasons to throttle, services like Netflix are in a poor position to do it correctly. The T-Mobile system that throttles users only when they’re approaching their data limits probably works better for most users because T-Mobile knows something that Netflix doesn’t know: how much data the user has purchased and where the user stands with respect to their quota. Without that knowledge, the best an edge service provider like Netflix can do is guess, and their odds of guessing right are no better than a coin toss.
So why would Netflix want to throttle streams when it has no idea whether customers really want their streams throttled? At a bare minimum, the company should have made the 600 Kbps one of several selectable limits. Netflix essentially admits this in saying that it’s looking into making the limit user-controllable at some future date. Given that the system has been in place for five years, one wonders what the holdup has been.
In a perfect world, video stream limits would be a matter to be negotiated by the carrier and the user, as T-Mobile has made them. The carrier can create a more optimal system by bringing the content provider into the loop as well, as the Binge On program does, but carriers who strive for optimality run the risk of violating OIO or at least raising the ire of the pressure groups who generally take the side of the content providers.
Hence, the OIO’s hard line ban on unreasonable throttling is counter productive: Throttling is sometimes useful, and when it is, it’s best done by the carrier with cooperation from both the user and the content source.
Who Can Throttle and Who Can’t?
The OIO binds carriers to compliance with its hard line rules by banning “unreasonable discrimination,” but completely exempts content services from them. Given that throttling is a fact of life over mobile networks with limited spectrum, the OIO ensures that it will be done by the parties with the least capability to do it correctly. This is a recipe for poor user experience on the Internet and low investment in infrastructure, so…way to go FCC!
In the FCC’s defense, if its authority over ISPs is questionable, nobody argues that it has the power to regulate content sources in any convincing way. This isn’t a question of the FCC’s will to do the wrong thing every time, it’s the way our nation’s communication laws are written. As long as the law depends on an arbitrary distinction between content and carriage, the unified nature of network service will be ignored and Internet-based businesses will be regulated in arbitrary and irrational ways. The FCC hasn’t banned throttling by Netflix because it can’t. I don’t believe it wants to, but to do that the FCC would need to find some Title I authority that it doesn’t believe it has.
So the way things stand is that the only firms that can throttle Internet traffic without fear of sanction are those that lack the information to manage traffic in the consumer’s best interests, with two exceptions: a) Technically knowledgeable users can, in principle, control their own throttling if they have the interest to do so and limit their Internet user to services that allow them the privilege; and b) ISPs can throttle in the name of “reasonable network management” if they’re willing to defend the practice in the FCC’s star chamber. So far none have expressed enthusiasm for this prospect.
Who Must Disclose Throttling?
This is a very interesting question because it’s so convoluted. The OIO requires ISPs to disclose all kinds of things, especially throttling but it requires nothing of content providers. But the FCC isn’t the cop on the beat for content firms, the FTC is. There’s nothing to prevent the FTC from requiring firms like Netflix to disclose their network management practices to customers.
There is good reason for such a requirement from the ISP perspective because the pressure groups have trained their followers to blame the ISPs until they’re proved harmless.
Requiring network management practices – or those that might be confused with network management – to be disclosed helps the overall structure of accountability in the Internet space, something that’s sorely lacking today. The mere fact that Netflix was able to keep its throttling secret for five or six years tells us that we don’t really know what’s happening in Internet management. How many people have blamed carriers or phones for poor quality video streams over the past six years because they had no idea what was really going on?
And of course these hidden practices did not exactly hurt Netflix and its allies in the net neutrality movement from pressure the White House into enacting regulations that upset the formerly level playing field in Internet regulation.
So the answer to the question I’ve just asked is that ISPs have to disclose for certain, and it’s up to the FTC to level the playing field holding making transit and content providers to the same standard.
The Double Standard
Throttlegate is another example of how comfortable non-engineers have become with applying different standards to the same practices depending on whether the practices in question are done by favored players or disfavored ones. We see exactly the same dynamics at work in the Throttlegate matter that we see in the FCC’s treatment of DNS and personal privacy. Before the DC Circuit, the FCC argued that DNS is network management when done by an ISP, but magic fairly dust when done by Google or any other non-ISP. In the case of privacy, the FCC proposes an opt-in system for ISPs who wish to collect web surfing metadata while web services operate under an opt-out regime. So it’s consistent with this inconsistency for the FCC to hold that throttling is presumptively bad and in need of disclosure when done by ISPs but none of its business when done by anybody else.
Perhaps something good can come of the friction these double standards have created: If Congress is paying attention, it might be a little more motivated to level the playing field through legislation. That seems to be the only way out.
* This sentence edited from the original version.