DC Circuit Passes the Baton to Congress

The long-awaited DC Circuit Court opinion in the challenge to the FCC’s deregulation of Internet service finally dropped yesterday, and it’s heavy reading. The opinion is nearly 200 pages long, including two concurrences and one partial concurrence and partial dissent. The prize in the Cracker Jack box was a discussion of the amicus brief I filed along with colleagues John Day, Tom Evslin, Shane Tews, and Martin Geddes.

The court reached the only possible conclusion on the issue of the regulatory classification of Internet service: the FCC is free to classify under Title I or Title II according to Supreme Court precedent. Obama appointees Millett and Wilkins didn’t like this, but they reluctantly followed the law while complaining about it in their concurrences.

The court reached the rather shocking conclusion that the Order didn’t provide sufficient justification for its blanket preemption of state “net neutrality” laws, such as the ones passed in California, Colorado, and Vermont. While the court’s legal reasoning may be correct in some narrow, legalistic respect, the FCC has never had to justify federal jurisdiction over interstate commerce before.

Parts of the opinion read like applications for a Supreme Court seats, something we see in DC Circuit opinions quite often. The court did get some of the technical facts correct, especially on the nature of DNS. But it stumbled on questions about network management and the nature of Internet service.

The Plain Facts About DNS Support the FCC

The FCC order relied, in part, on the fact that the DNS and caching parts of the bundle of services offered by ISPs to consumers are Information Services under the law. This is sound because they consist of “generating, acquiring, storing, transforming, processing, retrieving, utilizing, or making available information.” When your apps retrieve IP addresses, domain names, or authentication keys from DNS, they are using an ISP-provided database.

Petitioners claimed, falsely, that DNS is an inextricable part of the ISP transmission service, but this claim was refuted by their own amici, professors Scott Jordan and Jon Peha, on pages 29-30 of the opinion:

“Internet architects deliberately created DNS to be entirely independent from the IP packet transfer function,” Jordan/Peha Amicus Br. 17, and “a BIAS provider’s DNS is an extraneous capability * * * not required for the core service,” id. at 17–18 (emphasis added). But if DNS is “extraneous” to operating the network, it is at least debatable whether DNS is used in “the management, control, or operation of a telecommunications system or the management of a telecommunications service.”

Amici for the Commission make related points, observing that “[a]n app’s DNS translation transaction ends before the BIAS transmission begins,” “DNS transactions do not provide the BIAS provider with information about the best path to the destination,” and they “do not have the power to either optimize or impair the BIAS provider network.” Bennett et al., Amicus Br. 13.

Thus it is at least reasonable not to view DNS as a network management tool. Id. at 13–14. Granted, Jordan and Peha remark that running DNS helps an ISP “reduce[] the volume of DNS queries passing through its network.” Jordan/Peha Amicus Br. 18. But in the deferential posture of Chevron the points quoted above by Jordan/Peha seem in part to support the Commission’s reading of the record (consistent with Bennett et al.) as showing that, whereas “little or nothing in the DNS look-up process is designed to help an ISP ‘manage’ its network,” 2018 Order ¶ 36, DNS is “essential to providing Internet access for the ordinary consumer,” id., for whom “DNS is a must,” id. ¶ 34 (quoting Brand X, 545 U.S. at 999).

In fact, the Internet didn’t even have a DNS for its first decade of operation and nothing about DNS relates to network management. DNS exists for the storage and retrieval of information that enhances the user experience of the Internet as a whole and has no bearing on the operation of the transmission pipes.

Not only is it reasonable to view DNS as something other than a network management tool, it’s the only reasonable position. An overly-broad interpretation of the telecommunications management exception (TME) was one of the Wheeler Title II orders principle defects.

The Title II Order Got This All Wrong

The Restoring Internet Freedom order was written in 2017 to correct the abundant errors in the Wheeler FCC’s 2015 Title II order. The Title II order was written in response to an infamous YouTube message from President Obama to the Commission: “reclassify Internet service under Title II of the Telecommunications Act.”

The opinion observes that the Title II order claimed that DNS is a network management tool inside the TME:

[The RIF order] observes that the Title II Order had essentially proceeded in a contrary manner, finding that the management-centered functionality of DNS predominated, so as to render it TME-worthy. “Although confronted with claims that DNS is, in significant part, designed to be useful to endusers rather than providers, the Title II Order nonetheless decided that it fell within the [TME].” Id. ¶ 38 (emphasis added). The Commission reasonably declined to follow this route (partly, as we shall see below, because it believed that it would cause the exception to swallow the rule in ways antithetical to its reading of Commission precedent and the Act’s goals). It chose a different, and reasonable, alternative.

But the Title II order’s analysis of DNS relied on legal rather than technical opinion:

366. DNS Falls Within the Telecommunications Systems Management Exception to the Definition of Information Services. As the Supreme Court spotlighted in Brand X, the Commission predicated its prior conclusion that cable modem service was an integrated information service at least in part on the view that it “transmits data only in connection with the further processing of information.”1024 That was so, under the theory of the Cable Modem Declaratory Ruling, because “[a] user cannot reach a third-party’s Web site without DNS, which (among other things) matches the Web site address the end user types into his browser (or ‘clicks’ on with his mouse) with the IP address of the Web page’s host server.”1025 The Commission had assumed without analysis that DNS, when provided with Internet access service, is an information service. The Commission credited record evidence that DNS “enable[s] routing” and that “[w]ithout this service, Internet access would be impractical for most users.”1026

In his Brand X dissent, however, Justice Scalia correctly observed that DNS “is scarcely more than routing information, which is expressly excluded from the definition of ‘information service’” by the telecommunications systems management exception set out in the last clause of section 3(24) of the Act.1027 Thus, in his view, such functions cannot be relied upon to convert what otherwise would be a telecommunications service into an information service. Therefore, consideration of whether DNS service falls within the telecommunications systems management exception could have been determinative in the Court’s outcome in Brand X, had it considered the question.

The Wheeler FCC had dozens of technologists on staff, including Scott Jordan. But instead of relying on expert technical judgment about the nature of DNS, it pandered to the court by repeating an offhand claim made by a judge in a dissenting opinion. In fact, Justice Scalia was wrong: routing in the Internet is a technical operation performed on the numerical addresses in IP packets, not on domain names.

It is aided by routing information propagated by Border Gateway Protocol (BGP). Internet routing doesn’t know whether the application got the IP address from DNS or whether it was pre-recorded. Routing doesn’t see DNS, doesn’t interact with DNS, and doesn’t require DNS. As our amicus said: “An app’s DNS translation transaction ends before the BIAS transmission begins.”

This is what the legal system does with facts all too often. A false statement by a judge in a dissenting opinion plays the role of a legitimate fact when repeated often enough. What did Justice Scalia really know about the Internet? Why would an order by an expert agency cite such a view and rely on it in such a crucial fact-finding exercise? The Wheeler FCC made the telecommunications management exception swallow any and all Information Services offered by ISPs.

Judge Millett’s Erroneous View of the Internet

Judge Millett’s concurrence is a bit, well, unhinged (to use her language.) First, she uncritically regurgitates plaintiffs’ claims that the broadband service of the Brand X era was a “walled garden” dominated by ISP-provided content whereas today’s Internet is radically different:

Brand X faced a “walled garden” reality, in which broadband was valued not merely as a means to access third-party content, but also for its bundling of then-nascent information services like private email, user newsgroups, and personal webpage development. Today, none of those add-ons occupy the significance that they used to. Now it is impossible “to deny [the] dominance of [third-party content] in the broadband experience.” USTA, 825 F.3d at 698.

“[C]onsumers use broadband principally to access third-party content, not [ISP-provided] email and other add-on applications.” Id. (emphasis added). In a nutshell, a speedy pathway to content is what consumers value. It is what broadband providers advertise and compete over. And so, under any natural reading of the statute, the technological mechanism for accessing third-party content is what broadband providers “offer.”

This confuses the dial-up Internet with the early broadband Internet. We never needed broadband to use walled gardens such as AOL; their trade was low-bandwidth chat and proprietary content in text format. We moved from dial-up to broadband with the advent of the Web, a system of third party content.

The Ascendancy of the Web

The Web achieved ascendancy side-by-side with broadband because the Web’s rich content – graphics and audio streaming followed by video streaming – required broadband capacity. By the time Brand X was decided in 2005, the Web was utterly dominant. ISP portals provided by Excite @Home were search engines with a minimal supply of proprietary content, training wheels while users learned to navigate the Web.

It’s no accident that Mozilla’s account of this “walled garden with no actual walls” is similarly flawed. It utterly fails to acknowledge that the Excite service was a search engine for finding third party content. While the Web is even richer today than it was in the beginning of the broadband era, it was always the main attraction.

The nutshell view that Internet users are solely interested in content suggests that Millett is a “Netflix and chill” user. For most of us, the Internet is way to communicate with others through Facebook and Skype. A big and growing part of its appeal is access to remote processing services enabling stock market analysis, general gaming, and augmented reality.

The Internet of Today Is About Distributed Computing

The Internet of today is the apps on our personal devices communicating with partner apps on remote servers. It’s the system of distributed inter-process communications we described in the Amicus:

This communication that takes place within computer operating systems and through the BIAS network’s network operating system is known as inter-process communication, or IPC.  See John Day, Ibrahim Matta, and Karim Mattar, Networking is IPC: A Guiding Principle to a Better Internet, CoNEXT ’08, Proceedings of the 2008 ACM CoNEXT Conference at 1-3 (2008).  The distributed applications engaged in IPC do not necessarily know whether their partner application is in the same device or across a network, because IPC functions in the same way in either case: it notifies the sending partner that its message has been sent, it delivers a signal when the message is received, and then delivers a correlated reply to the appropriate process.

Sophisticated IPC systems prioritize signals and messages according to their relative urgency and tailor their behavior according to application needs.  In the BIAS network, IPC’s role is to analyze, interact with, and react to the information it receives, not just serve a transmission function.  Through their inherent packet multiplexing capability, Internet service providers support multiple concurrent IPC sessions to distinct IP endpoints and even to distinct processes within the same endpoint; multiplexing is outside the user’s control on the PSTN.  And just as IPC is not a transmission service when it takes place within a single device, it does not become one when performed over a larger network.

Understanding that the Internet is an IPC platform is the key to advancing from a telecom-centric view of the system that leaves so many important questions unanswered to an Internet-centric model that allows us to make sense of the interactions we see. The telecom-oriented view of the net neutrality advocates tells us that the only bad actors we need to worry about are ISPs, but in the overall sweep of today’s Internet concerns ISPs don’t move the needle.

The IPC Model Has Long Been on the Brink of Emergence

The Brand X decision made the very cogent observation that all the services offered by ISPs (not just DNS, caching, email, and personal webpages) are information services because they’re about Internet access:

…the high-speed transmission used to provide cable modem service is a functionally integrated component of Internet service because it transmits data only in connection with the further processing of information and is necessary to provide such service. The Commission’s construction therefore was more limited than respondents assume.

When an end user accesses a third party’s Web site, the Commission concluded, he is equally using the information service provided by the cable company as when he accesses that company’s own Web site, its e-mail service, or his personal webpage. As the Commission recognized, the service that Internet access providers offer the public is Internet access, not a transparent ability (from the end-user’s perspective) to transmit information.

Emphasis added. Historical spin has transformed the FCC’s arguments in Brand X into the walled garden/personal web page simplifications offered by Judge Millett, Mozilla, and the Title II order; but they never were that. In fact, the FCC argued that broadband Internet service inherently provides transmission only in connection with the further processing of information. The RIF order makes a similar observation at ¶ 27:

Below, we first explore the meaning of the “capability” contemplated in the statutory definition of “information service,” and find that broadband Internet access service provides consumers the “capability” to engage in all of the information processes listed in the information service definition. We also find that broadband Internet access service likewise provides information processing functionalities itself, such as DNS and caching, which satisfy the capabilities set forth in the information service definition. We then address what “capabilities” we believe are being “offered” by ISPs, and whether these are reasonably viewed as separate from or inextricably intertwined with transmission, and find that broadband Internet access service offerings inextricably intertwine these information processing capabilities with transmission.

Unlike telecommunications that aims to bounce sound waves around in our ears, broadband Internet service aims to move information from one process (or processor) to another, inherently a different proposition that demands its own norms, laws, and regulations. While some complain that this construction enables information services to swallow transmission, they have no problem with the ridiculous proposition that transmission must necessarily swallow information services.

Get real, the Internet is all about the processing of information and the historical distinction between content and carriage is untenable. Moving information is an act of processing in today’s world.

Congress Needs Help

The main takeaway from the DC Circuit’s patchwork quilt of an opinion is that the law is seriously confused about the Internet. We can also draw that conclusion from the complaints and defenses of Section 230 and the ongoing cybersecurity debacles we face on a variety of services. (Apple’s recent software updates have been plagued by security bugs that have caused further updates and updates to updates, for example.)

The Internet is close to being too complicated to operate, let alone to use. Trying to stuff Internet service into rusty regulatory buckets hasn’t worked because it’s not congruent with technical realities.

We need to have a big discussion about the obligations of information platforms that doesn’t drive a wedge between network operators, social networks, device and search companies, and other providers of information services. I don’t think Congress is ready to regulate any of these things, but it needs to admit we have a problem and commit do the background work on solving it.