I want to thank Eugene again for inviting me to guest-blog last week about my new article, The Five Internet Rights. See Parts I, II, III, IV, and V. I thought I’d follow up with just one more post to respond to some of the reader comments I received on the series.
In the first place, it was certainly an interesting (and educational) experience trying to distill a 100-page, heavily footnoted, academic article into five short-form blog posts. That translation forced me to elide over many important details and caveats that were addressed in the article, some of which readers focused on. For example:
On the technical side, the savvy DNS practitioner might point out that having one’s domain name suspended doesn’t amount to getting kicked off the internet (just ask gab.com and ar15.com). Often, when a registrar suspends a domain name, it gives the registrant an opportunity to find another registrar. Given that more than a thousand ICANN-accredited registrars currently offer their services to the public, a registrant faced with a suspension for purely ideological reasons can usually find a substitute. Or a truly beleaguered speaker could theoretically become her own registrar (with a fair amount of money and effort). And even if a top-level domain registry (as distinct from a registrar) permanently deprived you of your domain name (say, example.com), you could probably register another string in a different top-level domain (say, example.xyz or even example.ru).
Likewise, even if a regional internet registry (RIR) revoked your IP addresses, you might manage to procure substitute address space within another RIR’s service region. I address these nuances in a section of the article that distinguishes between “strict” and “fuzzy” classes of intermediaries and notes the hierarchy of control from registrars to registries to ICANN. But obviously, that would have been too in the weeds for a blog series (as would have been any description of the seven-layer OSI network stack).
Another important clarification is that even if a person lost her domain name, IP address, and commercial internet connection, those actions would only kick her website off the internet; they wouldn’t necessarily kick her off the internet. She could probably continue to leverage the internet to express her viewpoints using non-web technologies, such as email, FTP, peer-to-peer, or video conferencing. I acknowledge as much in the article. But I also argue that websites enjoy a special status when it comes to online expression. Websites alone combine control, social valence, accessibility, discoverability, authority, and permanency. Not so with spoof-able emails, undiscoverable file objects, or ephemeral Zoom-casts. A viewpoint that is denied a home on every website might not be banished from the internet entirely, but it would be effectively banished from the public internet.
Some readers focused on certain anecdotes I used to illustrate the evolution of content moderation as it has progressed down the internet stack. Did GoDaddy suspend ar15.com simply because GoDaddy disliked lawfully expressed viewpoints on the site, or did GoDaddy find instances of actual (unlawful) incitement to violence? Did LACNIC revoke IP addresses used by Parler because it wanted to take the unpopular social media platform offline, or was it simply enforcing neutral policies?
As I address in the article, instances of deplatforming will often be messy or ambiguous. Take LACNIC, for example, which supposedly revoked IP addresses belonging to DDoS-Guard (Parler’s cloud hosting provider) because DDoS-Guard had used a shell company in Belize to obtain the addresses. Yet corporations often create local subsidiaries (with no employees) in foreign jurisdictions solely to obtain licenses or other resources granted only to local entities. And as any telecom lawyer who has incorporated local subsidiaries in different countries to obtain telecom licenses for a global OTT service will tell you, that is a perfectly acceptable practice. Moreover, the provision at issue, § 1.14 of the LACNIC Policy Manual, requires only that address holders be “legally constituted within [the LACNIC] service region”; it contains no requirements as to a minimum of employees.
Perhaps DDoS-Guard lost its IP addresses because it did not use them primarily to serve networks in Latin America, which is also contemplated by LACNIC policy. If DDoS-Guard had violated that principle (and I’ve seen no reporting stating that it had), then that might indeed have been a viewpoint-neutral reason for the revocation. But the fact that many address holders are likely in a similar position highlights another problem: the opportunity for selective enforcement. If there is any desire to see an unpopular speaker deplatformed, it will often be possible to find some technical violation. Many intermediaries prohibit “objectionable” content (which is usually not defined), GoDaddy may cancel a domain name if it receives an “excessive amount of complaints” from the public, and RIPE, another RIR, reserves the right to revoke IP addresses from any holder that merely “cause[s] damage” to its name, which theoretically could be interpreted to encompass situations where enough people criticize RIPE for enabling an unpopular site to stay online. As one commentator noted, in 2021, GoFundMe demonetized the Canadian trucker protesters based on only three instances of minor illegality while continuing to fund protests in Portland that “set fire to police stations, vandalized city hall, wielded weapons and injured police officers.”
And just as concerning, I would argue, was Ukraine’s attempt to revoke Russia’s IP addresses and top-level domains. That attempt perhaps serves as a cleaner example of the growing interest in using internet architecture as a tactical weapon (not to take anything away from Ukraine’s cause, which I otherwise support).
In any event, and circling back to my opening remarks, it’s important to understand that The Five Internet Rights is ultimately a theoretical piece. Its point is not to demonize any particular intermediaries or to claim to know their motives. And its thesis doesn’t depend on any particular instances of deplatforming or my interpretation of them. Rather, it offers a theory of interventionism—an answer to the thorny question of when, if ever, the state should intervene in private content moderation.
Doesn’t your thesis require you to support net neutrality?
For the most part, yes. My thesis is that an intermediary’s “content moderation” practices—its decision to revoke a resource used by a third party to publish lawful content on the internet—should be subject to regulation (in the U.S.) if (and perhaps only if) the revocation of that resource presents the risk of viewpoint foreclosure. That is, if (1) that resource class is essential to operating a public website and (2) a person deprived of that resource cannot realistically create a substitute. By my reckoning, three resource classes satisfy this test: networks, IP addresses, and domain names. The “five internet rights” call for non-discrimination protections across these three classes and only these three classes.
That thesis is bound to upset some folks on both the left and the right (no one loves a moderate). Conservatives might complain that it doesn’t touch social media (or any other websites), search engines, app stores, or even cloud computing (inclusive of content delivery networks and DDoS mitigation services). Google would remain free to de-index 4chan, Apple could keep Parler out of the App Store, and Amazon could boot Parler off AWS. After all, none of those actions would take the targeted site permanently offline. By contrast, progressives might oppose any policy that would keep the worst kinds of content online, even if that online presence is limited to backwater websites that few visit or link to.
Net neutrality (before it was repealed) guaranteed only a right to “accessibility”—the right not to have one’s users blocked from accessing her lawful website by their ISPs (their last-mile networks). But it didn’t protect a website operator’s ability to connect her website to the internet in the first place, to maintain a static IP address and a resolvable domain name, or to have her website’s packets faithfully routed through intermediate backbone networks. For those, you need additional rights of connectivity, addressability, nameability, and routability.
Some have told me that they think of the five internet rights as a kind of expanded net neutrality (perhaps it might be called “internet neutrality” or “infrastructure neutrality”). That’s true in a sense, but there’s an important distinction. From the beginning, the concerns that motivated the political left to pursue net neutrality were economic in nature. The left wanted to prevent ISPs from leveraging their networks and their access to subscribers (in telecom parlance, their “terminating access monopolies”) to extract rents from websites in the form of tolls, paid prioritization, or zero-rating or to advantage their OTT services over those of competitors. By contrast, the concerns that have motivated the political right to pass laws like those in Texas and Florida in recent years have been moral in nature (in the deontological sense and not to make any value judgments about the moral rightness of those efforts). The right has wanted to prevent websites like social media platforms from discriminating against users for ideological reasons. Both are non-discrimination enterprises, but their theoretical foundations differ.
What the right and the left should appreciate is that their interests now align. The FCC’s 2015 Open Internet Order prevented ISPs from blocking their subscribers’ access to lawful websites and applications. That broad prohibition wasn’t conditioned on whether an ISP might block a website for economic reasons (to charge a toll or disadvantage a competitor) or for moral reasons (to counter the viewpoints expressed on the site). If the right is concerned about private “censorship” on the internet, it should consider giving the left the economic neutrality it has long desired in exchange for the ideological neutrality the right now craves. Given that the ISP ecosystem already abides by net neutrality principles (whether because of state net neutrality laws, the EU Open Internet Regulation, or market forces), that seems like a small concession. Ideally, the left and the right could agree on a broad neutrality framework that encompasses all aspects of the network (beyond net neutrality’s focus on last-mile, mass market access), as well as DNS and the IP address system, and that protects against both economic and ideological discrimination. (Precisely what the “five internet rights” are designed to do.)
At least, that seems like a sensible compromise to me.