Back to

Why Consensus Matters:
The Theory Underlying ICANN’s Mandate to Set Policy Standards for the Domain Name System

by David R. Johnson and Susan P. Crawford[1]
Webposted on 23 August 2000

 We have argued that ICANN needs to return to its roots as a consensus-based organization, paying close attention to the limits on its powers to enforce its collective decisions.[2] In this article, we explain why it is a good idea to require documented consensus (showing widespread support, and no strong or reasoned opposition, among the community of those directly affected by a proposed policy) as a condition to requiring compliance with ICANN’s policy standards. We also elaborate on why certain decisions by ICANN -- like the entry into new contracts that expand its “consensus policy” regime to new contractual partners in an equitable way -- should not be required to be supported by documented consensus. 

We all think we know the virtues of democracy -- and that it's "the worst form of government except for everything else." We're wrong about that last part in the context of the global internet. The democratic formula used in the 18th century for the United States does not work (and will not work) for the internet or for ICANN. 

Democracies are characterized by majority voting rules, legislators who are supposed to represent the body of people affected by their decisions, one-person-one-vote rules, attempts to ensure that the group of those with a right to vote largely overlap the group of those required to comply with the resulting rules, and "subsidiarity" rules that create a hierarchy among governmental actors creating particular sets of rules (keeping the units of the state from creating rules that will, for example, impede commerce among the several states). By definition, majority voting systematically disregards the interests of minorities. For that reason, the founding fathers of the American form of democracy went to great lengths to set up checks and balances, to establish courts and basic constitutional rights, and to arrange for deliberative legislatures. Federalist 10 explains that careful selection of the size of represented territories -- and care with respect to the size of the legislative body itself -- is needed to assure a deliberative process that will result in policies that best serve the overall public interest. 

One-person-one-vote rules cannot readily be policed on the internet. Even if they could be, we simply cannot assure that those with a right to vote will largely overlap the group of those required to comply with the resulting rules -- something the geography of territorial sovereigns has previously tended to do. No matter how open and transparent the process used to create the ICANN Board and even the ICANN At Large Membership itself (and there is some question about that), these bodies cannot credibly claim to be "representative" of the body of those impacted by and/or required to follow ICANN's rules. Indeed, the set of impacted parties will vary from rule to rule, and the nature and size of the impacted group will evolve as participation in the internet develops. We cannot realistically expect everyone with some stake in the internet domain system to join ICANN and participate actively. Nor is there any good reason to believe that the tiny percentage of potential stakeholders who do join and vote will be "representative" of those who are impacted. (While it is true that territorial democracies often have low voter participation at the polls, at least a low turnout in a particular county's sheriff election doesn't produce a result allowing the elected sheriff to arrest anyone outside the county! And you can always leave the county. This is not true when the "jurisdiction" for an election is defined as the internet itself.) 

More importantly, insofar as our best democracies are designed to promote not just voting by citizens but also constructive and thoughtful dialogue in appropriately-sized legislative chambers populated by civic-minded, non-corrupt, legislators (an ideal rarely realized), that ideal does not alone work to produce fair results. The rights of minorities must be protected -- but in the context of ICANN, where is the "Bill of Rights" that could establish baseline protections for those against whom some majority (or supermajority) within the ICANN structure might combine? Where are the "checks and balances" in the ICANN structure, given that, as a private corporation, ICANN's Board may amend its bylaws at will? (It would be a bad idea to write a Bill of Rights, by the way, since such a document would imply the existence of individual rights as against a sovereign state -- and we should not be making ICANN into a sovereign.) 

Finally, even under the best conditions of participation, a legislative body elected "democratically" by those with a stake in the internet domain name system would not have any right to speak as a sovereign on behalf of its "netizens." Each participant will still be governed by local, territorially-defined law. No one is prepared to create a world government -- even a democratic one. Even if we could apply democratic "representation" principles perfectly, we would not have created a sovereign that had the last word on what law applies to all netizens. The resulting ICANN would have no power to enforce its decisions other than by contract. And there is every prospect that the registries and registrars with whom ICANN contracts would not voluntarily agree to contracts that submit decision-making to voting by an unpredictable populace of those who may or may not have a significant stake in (or even pay much attention to) the resulting rules. 

Although 18th century democratic concepts do not make sense for ICANN, the required demonstration of consensus support for enforcement of any ICANN policy does make sense -- because it serves the same basic goals pursued in physical, territorial contexts by democracies. ICANN's goal should be to find policies that best serve the broadest group of internet participants and disserve the smallest possible group (and only to an extent such that the opposing group can grudgingly go along with such policies in order to get something done). In a context in which democratic "representation" is impossible, the key virtue of consensus is that it identifies uniform solutions that have widespread support and no substantial opposition, thereby protecting minorities against oppression, allocating decisions efficiently among levels of policy-makers, and facilitating constructive deliberation. 

The test for consensus is instinctively known by all of us (and is set forth in ICANN's contractual agreements): opposition to a particular policy is limited in scope and intensity (or is unreasoned), and opposition does not stem from those specially impacted by the policy. Consensus does not mean unanimity or even that a particular percentage of those voting have agreed to a particular policy. Rather, it means that there is very substantial support for the policy and very good reason to override any remaining dissent. Having a written report demonstrating the presence of consensus is crucial (and is contractually required of ICANN). Such a report will have to show that there is a need for uniform resolution of the issue in question. It will also need to demonstrate the existence of real consensus, which will require an appropriate effort to survey the views of those impacted and an analysis of the nature and intensity of support and opposition. (It will not be enough just to show that the process was open for comments and no one showed up.) Only with such a written record would the ICANN Board be able reasonably to decide that any remaining opposition should be overridden. 

We believe that the requirement of documented consensus will produce better results than any process of democratic elections and deliberation among "representatives," for the simple reason that those whose views will be consulted in the dialogue required to document consensus will be closest to the problem and to the solution. Impacted parties, appropriately surveyed, will be able more accurately to reflect their own values and needs and views than any set of “representatives” could be. “Democratic” elections and deliberations will “get it wrong” a good portion of the time. In contrast, a process that requires production of a persuasive report demonstrating that those substantially impacted by a policy generally support it -- and that very few impacted parties oppose it both strongly and rationally -- will probably “get it right.” 

The process of consensus-building (and producing a thorough written record and report) is not easy and will be subject to subversion. The presence of a “consensus” can be suggested when nothing of the kind has been achieved. It can even be claimed to have occurred within a Board or a Names Council or some other body (which will then turn out not to have been truly representative of impacted parties). But the ICANN standard does not talk about “consensus on the Board.” (Indeed, the contractual grant to the Board of special temporary emergency powers proves the rule that the Board cannot make top-down law under any other, non-emergency, circumstances.)[3] The standard embodied in ICANN’s contracts refers to a documented consensus among those impacted by a policy. It is not yet clear precisely how best to go about making the demonstration of consensus required of ICANN. (It has hardly been attempted.) One thing is clear: the fact that even a “democratically” elected set of ICANN representatives voted (by any level of majority) for some policy has little to do with the question whether those impacted by it can credibly be claimed to support (or not vigorously oppose) it. ICANN simply cannot offer a plausible argument that it has received powers by means of delegation to it, as a centralized legislature, from the relevant, impacted groups. 

Of course, no one can expect all of the impacted parties to get involved. But a “representative sample” of views following actual notice and detailed dialogue (note: quite a different sense of “representative” from the traditional democratic understanding) can be used to demonstrate that there is widespread support and little strong opposition to some proposed policy standard. Working groups and deliberative polling can be used along with outreach to particular parties who will be required to implement a proposed policy. (It would also help to use the net itself to gather views about any proposed policy.) 

If a significantly impacted group strongly objected to a proposal, the “democratic” process would call for a vote. The “consensus” process, by contrast, calls for a discussion and the development of additional evidence regarding the actual distribution of views among those impacted -- and, often, for the evolution of proposed policies to some more readily acceptable alternative. To be sure, a vote is more certain to lead to a rapid decision. But such efficiency is an unalloyed virtue only when there is a need for a sovereign unambiguously to make the law for a territorial place. On the internet, the absence of a “demonstrated consensus” might well be the right result in the sense that it will leave decisions on contentious matters to be made locally. Additionally, strong opposition from an impacted group will have the effect of knocking ICANN's policy-making off a low (suboptimal) level in policy "fitness space" -- and may well force a dialogue that leads to higher ground. Surveying impacted groups (rather than elected "representatives") assures congruence between those who collectively make the rules and those who must abide by them. 

In America, political parties emerged as a method by which to organize the greatly-expanded electorate of the early 19th century. American parties cut across American class lines, are open to all, and provide forums for debating important issues and helping in the peaceful resolution of conflict. Given the vastly expanded "electorate" potentially affected by ICANN policies, and the difficulties inherent in doing the individualized work of achieving consensus, it may make sense for ICANN to borrow from democracies the idea of fostering the emergence of many open, bottom-up “parties.” These parties could be groups that share viewpoints. They could grow large enough to insist on sending a representative to voice their shared views in a working group or deliberative polling session. Unlike the current “constituencies” mandated top-down by ICANN within the DNSO, the purposes and labels and memberships of these groups would be open to change, so that new issues and new constellations of alliances could evolve. Their power to persuade other groups would stem from their power to collect and keep the support of committed members. The loyalty of such groups' “representatives” would be assured -- because, unlike membership in the internet at large, membership in a particular “party," however defined, would be optional. Because the representatives of the most powerful factions would not sit as a legislature passing top-down laws, and would only serve as one of many strong voices in the search for consensus, the dangers of capture by these factions would be mitigated. 

In a democracy, the question is “who’s got the votes?” (Even in the high-minded thinking of Federalist 10, the theory is, at best, “who is entitled to deliberate together on behalf of the people at large?”) In a consensus-based regime, the questions are very different: “What do we (who support a policy) have to do to convince you (who oppose it) to go along?” or “Is there enough agreement, and little enough opposition, that we should all (including the opposition) agree to override the opposition?” 

The “social contract” of democracy supposes that we are creating a leviathan, and that we must do so to provide legitimacy for the control of physical force within a given geographic territory. The “social contract” in a consensus-based regime is more of a peer-to-peer relationship (and may require more effort by the individuals involved). It asks each potential participant whether they will agree contractually to implement and abide by any future policy, sight unseen, provided that most people support it and those parties substantially impacted by the policy do not vigorously oppose it. It is a reasonable contract, given a certain minimum level of trust. It is also a real contract -- binding because of a signed document, not merely a philosopher’s musing about a hypothetical social act by one’s ancestors. 

On the INTER-net you are not part of the social contract unless and until you join. You can set up a private network. No rational private network would join an internet governed by an election-based democracy (unless it controlled the votes). But many private networks -- even though unsure of the possible impact of future policies -- might agree to be bound to go along with those future policies that are supported by similarly situated folks and not strongly opposed by someone who is particularly impacted (a someone who might turn out to be them!). Some registries, registrars, and registrants have signed exactly such contracts -- and by doing so have given ICANN all the power it has. 

The overall goal of the consensus process is to facilitate dialogue, not to maximize the number of rules generated. The consensus standard calls on all parties to suggest changes that, while not perhaps optimal, will persuade others not to oppose. It is a standard based on the ideal of cooperation (looking for the high ground, not the lowest common denominator that can marshal a majority) on which the net was built. 

This leads us to the reason why a decision by ICANN to expand this consensus regime through contracts with additional registries should not require demonstrated consensus support. The creation of new TLDs will have an effect on existing registries, registrars, registrants and various other parties like trademark owners, to be sure. But the addition of new TLDs will not require any of these actors to implement or abide by a new set of rules. (By contrast, ICANN’s bylaws might be violated if it were to give new registries a fundamentally and unfairly different deal. The whole point of the web of agreements that created ICANN was to provide equitable treatment to all who joined the ICANN consensus-based standard-setting regime.) The internet expanded because no newly connected networks needed to ask anyone's permission (other than a peer network) to use certain protocols. This prevented incumbent parties from stifling innovation to preserve an otherwise transient advantage. The ICANN policy standards can and should grow by similar means. 

Ultimately, ICANN is a contract-based regime. Properly implemented and limited, ICANN will facilitate a new form of peer-to-peer social contract. It will allow us to explore new ways to guide global online interactions. We should allow this bold new experiment to grow as long as there are new participants willing to sign up. They will do so if they know they are getting the basic ICANN deal: a reciprocal obligation on all participants to work together to find solutions that most support and few who are actually impacted vigorously oppose. They will not do so if they are asked to submit to world government or a stealth sovereign. 

Some will say that the whole idea of “consensus” is unduly idealistic, hopelessly fuzzy, or incapable of scaling to fit heterogenous groups. The IETF “hum” was fine, they will suggest, but could only work among a group of idealistic engineers. We believe the contrary is true. The engineers who gave us the net (hardly a non-contentious group) also gave us the first inkling of a better way to evolve policy in a global online space. It is 18th century democracy that will not (for now) scale. 

The presence of documented, demonstrable consensus will ultimately be judged (by the parties, the Independent Review Panel and, ultimately, courts) in a context-rich, case-by-case manner. The votes of the Names Council or Board in connection with proposed policies will become occasions for reasoned judgments as to whether consensus exists, not just for voting or logrolling among those lucky enough to claim seats on these steering committees. It will matter what the policy in question was, what the impacts were, what was done to reach out to impacted parties, and what the nature of supporting and opposing interests and arguments have been. But the contractual ICANN standard requiring a persuasive demonstration of consensus is far less vague -- and will lead to far more legitimate rules -- than an entirely unpersuasive claim that an elected Board (or Names Council) speaks as a democratic, "representative" government on behalf of the internet community as a whole. 

The goal of all who wish the ICANN experiment well is to enhance the legitimacy of ICANN decisions. That legitimacy will come from two sources. First, those required to comply with ICANN policies will voluntarily have entered into contracts that commit them to abide by consensus policies on appropriate topics. Second, the policies that are adopted and can be enforced will have been supported by detailed reports that show that such policies reflect very broad agreement and that opposition is either very limited, unreasoned, or comes from those who are not in fact adversely impacted. In contrast, an assertion that ICANN is “democratic” will provide the shakiest of foundations on which to support ICANN’s policy regime. We need to base ICANN's credibility on the consensus standard so that we can get on with a searching dialogue regarding which policies for the internet ought to be established on a centralized basis and which should be left for “local” and decentralized decision-making. 

[1] David R. Johnson is head of the EGroup at Wilmer, Cutler & Pickering. He has previously represented Network Solutions, Inc. Susan P. Crawford is also a partner with Wilmer, Cutler & Pickering. This article reflects only their personal views. 


[3] See, e.g., ICANN Registrar Accreditation Agreement, I(B)(4).