THE DRAFT FINAL REPORT OF THE GNSO NEW GTLDS COMMITTEE ON
THE INTRODUCTION OF NEW GENERIC TOP-LEVEL DOMAINS
June 6, 2007
Statement by Christine Haight Farley
Professor of Law, American University Washington College of Law
I want to begin by commending the GNSO New TLDs Committee on their policy recommendations and implementation guidelines for the introduction of new top-level domains. Through the Draft Final Report, ICANN has explicitly stated its intention to make the gTLD application process open and transparent. The Draft Final Report has focused the issues and prompted a useful discussion. However, because I believe that the Draft Final Report includes a number of misstatements of domestic and international trademark law, I offer my legal analysis of these provisions.
I will address my remarks only to Recommendations 2, 3 and 6 as these recommendations rely heavily on trademark law concepts.
Before I make observations specific to these recommendations, I would like to offer some general remarks about the overall incongruence between trademarks and domain names. It is important to note at the outset this general lack of equivalence between trademark law and domain name policy. For instance, trademark law the world over is fundamentally based on the concept of territoriality. Thus, trademark law seeks to protect regionally-based marks and market-based marks without implication for the protection or availability of that mark in another region. In contrast, domain names have global reach, are accessible everywhere and have implications for speech around the world.
Likewise, trademark protections hinge on what the relevant consumer thinks. Again, the law considers the viewpoints of consumers of a particular country, region or market, and acknowledges the variability of these viewpoints across regions. Domain names are not directed to a certain class or geographical region of consumers, they are accessible to all. Therefore in order to take account of consumers’ viewpoints, it would be necessary to consider a global public. The resulting one-size-fits-all approach would be anathema to trademark law in that it would leave consumers confused in one place while unjustifiably denying speech rights in another.
Finally, trademarks rights are not applied abstractly or in theory, but are always considered in context. Thus, in order to determine whether the use of a mark by another would likely cause confusion, it is necessary to analyze how the mark is used in commerce. Consideration will be given to what goods or services bear the mark, what overall design or color scheme accompanies the use of the mark, what the level of consumer sophistication is, what marketing channels are used, etc. GTLDs are necessarily abstract. We can not know in advance what the content of a website hosted on a certain domain will be. It is therefore impossible to make fine-tuned conclusions about the appropriateness of certain domains. For these reasons, I strongly urge domain name policy makers to consider carefully the appropriateness of importing trademark law concepts into domain name policy.
Recommendation 2: Strings must not be confusingly similar to an existing top-level domain.
In this recommendation, the Committee seems to be conflating two distinctly different concepts: confusing similarity and likelihood of confusion. The Draft Final Report states that ‘confusingly similar’ is used to mean that there is a likelihood of confusion. However, confusingly similar is a different legal standard than likelihood of confusion. The Committee appears to base this recommendation on Section 18.104.22.168 of the ICANN Registrar Accreditation Agreement, which it cites, implying that the legal standard is consistent. But that section of the ICANN Agreement explicitly employs the standard of infringement, which is likelihood of confusion.
A determination about whether use of a mark by another is ‘confusingly similar’ is simply a first step in the analysis of infringement. As the committee correctly notes, account will be taken of visual, phonetic and conceptual similarity. But this determination does not end the analysis. Delta Dental and Delta Airlines are confusingly similar, but are not likely to cause confusion, and therefore do not infringe. As U.S. trademark law clearly sets out, the standard for infringement is where the use of a mark is such is to be likely, when used on or in connection with the goods of such other person, to cause confusion, or to cause mistake, or to deceive. While it may be that most cases of confusing similarity are likely to cause confusion, because the infringement standard takes account of how the mark is used, some cases of confusing similarity will not likely cause confusion.
In trademark law, where there is confusing similarity and the mark is used on similar goods or services, a likelihood of confusion will usually be found. European trademark law recognizes this point perhaps more readily than U.S. trademark law. As a result, sometimes ‘confusingly similar’ is used as shorthand for ‘likelihood of confusion.’ However, these concepts must remain distinct in domain name policy where there is no opportunity to consider how the mark is being used. As applied to domain names, the only level of analysis is the first level of analysis: confusing similarity.
A related problem with this recommendation is that it equates domain names with trademarks as legally protectable properties. They are not. Trademarks are legally protected intellectual property because it is believed that the commercial use of a mark by another that is likely to cause confusion would injure consumers. Trademarks are legally protectable intellectual property also because their owners have developed valuable goodwill in the marks. Neither of these conditions of legal protection apply in the case of domain names.
Moreover, it is not clear what consumers would be confused about when encountering a string that is confusingly similar to an existing top-level domain. Because, unlike trademarks, strings are not inherently commercial communication means, it does not follow that consumers would incorrectly assume that the string would indicate source of origin.
For example, http://nmhm.washingtondc.museum/ does not suggest to consumers a connection with www.museum.com.
Beyond top-level domains, the Draft Final Report states that “strings should not be confusingly similar either to existing top-level domains like .com and .net or to existing trademark and famous names.” The Draft Final Report notes that the Committee relied on a wide variety of existing law to reach this standard. And yet, ‘famous names’ is not a legal category under any trademark law. International trademark law grants rights to well-known marks and to ‘trade names,’ and U.S. law grants rights to famous marks, but ‘famous names’ seems to be a construct created by the Committee. Clearly, the domain name policy should protect only recognized intellectual property.
Recommendation 3: Strings must not infringe the existing legal rights of others that are recognized or enforced under generally accepted and internationally recognized principles of law.
There are simply too many legally recognized trademarks in the world to make this recommendation workable. The United States alone registers well over 100,000 trademarks each year and there were 1,322,155 active certificates of registration last year. In the United States, state registered trademarks and common law trademarks are also legally recognized. Protected trademarks include generic terms, geographical terms, names, and fanciful words.
Recommendation 6: Strings must not be contrary to generally accepted legal norms relating to morality and public order.
The Committee is correct that a variety of trademark legislation restricts trademark registration based on some notion of offense or immorality. Unfortunately, the Committee seeks to extend this trademark law concept to domain name policy. This extension is not a natural one and presents many problems in its application.
Where these content restrictions exist in trademark law they are understood as merely restricting the registration of trademarks, not the use of such trademarks. That is, under certain legislation a trademark may be deemed un-registrable but the trademark owner may still use the trademark in commerce and may even have the benefit of legal protection over the trademark. The only restriction is that the trademark owner is denied certain benefits of registration.
The United States has such a content restriction in its trademark law. What saves this legislation from violating the First Amendment is that it is not a restriction on use; it is merely a restriction on certain legal benefits deriving from federal registration. Any restriction of the use of the trademark would need to comply with the First Amendment. For instance, a mark may be restricted from use where it has been found to be obscene. Obscenity is a legal category whose threshold is well above the category of immoral or offensive speech.
The restriction of a generic top-level domain is more akin to the restriction on use than to the restriction on federal trademark registration. Because restricting offensive words in gTLDs would concomitantly restrict the ability of all speakers, commercial and non-commercial, ICANN should consider legal models outside of trademark law that better address the balance of speech rights.
This recommendation also illustrates the lack of fit between trademark law and domain name policy. Because trademark law is territorial in nature, legal standards reflect the consumer perspectives of the particular state. These standards are thus expected to vary from state to state as the way consumers respond to marks in one state may differ from the way consumers would respond to the same mark in another state. Trademark content restrictions are similar in approach. For instance, under U.S. trademark law, a mark will be refused registration if it is deemed to be scandalous or immoral when considered from the perspective of a “substantial composite of the general public.” The ‘public’ is understood to mean the U.S. public. In order to extend this legal standard to domain names it would be necessary to consider a substantial composite of the general public of the entire world. This is obviously an unworkable standard.
Moreover, trademark law standards are always applied in the context of how a consumer would encounter the mark. Thus, the USPTO and the courts consider the entire label, what products or services are sold under the mark and what channels of commerce and marketing will be used. As a result, marks challenged as being scandalous may in fact be found to have a double entendre. The extension of this trademark law standard to domain name policy thus risks prohibiting words as generic top-level domains that could well be used in inoffensive ways.
A few other observations are in order. First, under U.S. trademark law, in cases of doubt or ambiguity, both the USPTO and the Federal Circuit will pass the mark to publication to give others the opportunity to object. If ICANN finally decides to employ any content restrictions, erring on the side of permitting the speech should be the rule.
Second, the Paris Convention permits rather than requires content restrictions. Article 6s of the Paris Convention merely allows a Member state to deny registration to a mark duly registered in another Member state on the grounds of morality or public order. This article makes clear the expectation that a mark may be acceptable in one state, while it is offensive in another. The WTO TRIPS Agreement is silent on content restrictions.
Finally, the provision in U.S. trademark law that prohibits registration of scandalous or immoral marks is little known or utilized. Furthermore, the majority of challenges brought under this provision are brought by third parties and not the USPTO.
Christine Haight Farley
Professor of Law and Associate Dean for Faculty and Academic Affairs
American University Washington College of Law
Washington, DC 20016