Analysis and Suggestions Regarding NSI Domain Name Trademark Dispute Policy

by Carl Oppedahl(1)

[Note to the reader: This is a draft dated 8 September 1996 of a paper presented on September 9, 1996 by the author for the Information Infrastructure Project at the conference "Coordination and Administration of the Internet" held at Harvard's Kennedy School of Government on September 9 and 10, 1996. The final version of the paper appears in Coordinating the Internet published by MIT Press as part of the Information Infrastructure Project program.] A version of the paper also appears in the Fordham Intellectual Property, Media & Entertainment Law Journal, vol. 7, no. 1, Autumn 1996, p. 73.

In Luna in 2075 phone numbers were punched in, not voice-coded, and numbers were Roman alphabet. Pay for it and have your firm name in ten letters -- good advertising. Pay smaller bonus and get a spell sound, easy to remember. Pay minimum and get arbitrary string of letters. ... I asked Mike for such a ... number. "It's a shame we can't list you as ?Mike.'"

"In service," he answered. "MIKESGRILL, Novy Leningrad. MIKEANDLIL, Luna City. MIKESSUITS, Tycho Under. MIKES--"

(Heinlein, Robert A., The Moon is a Harsh Mistress, 1966 )(2)

Heinlein not only wrote of something like the Internet's domain name system twenty years before its time, he also wrote of what happens if the domain name you want is already taken. MIT registered its domain name in May of 1985, and IBM registered its domain name in March of 1986. But it wasn't until about 1993, when the World Wide Web brought the Internet to every computer screen, that domain names got to be a big deal. And it wasn't until about 1995 that fights over domain names began to hit the news.

This chapter talks about how certain types of trademark disputes on the Internet are handled now, and offers suggestions as to how they might be handled in the future. At present nearly all Internet domain names are administered by Network Solutions Inc. ("NSI"), a contractor for the US government, so NSI's policies are very important and will be discussed from several perspectives in this chapter, although the recommendations made here are of general applicability to other domain name registration authorities and to proposed new top-level domains.

This chapter assumes that the reader is knowledgeable about the Internet and the World Wide Web and with the system of domain names used on the Internet, and assumes further that the reader has some general familiarity with the law of trademarks. A number of excellent papers(3) set forth basic trademark principles and their connection with the Internet.

Domain names seem to be special

There are many ways in which someone on the Internet could trigger the ire of an owner of a trademark or some other form of intellectual property. A web site could contain someone's registered trademark. A web site could pluck an image (a trademark, or a Dilbert cartoon) from some other site and incorporate the image into its own web page. A web site could contain material protected by copyright that has been copied from elsewhere without the permission of the copyright owner. A domain name could be similar to (but not identical to) some trademark. A third-level domain name (e.g. could be identical to a famous domain name. Or a second-level domain name (e.g. could be identical to some trademark.

Each scenario has arisen many times in recent years, and in all cases but one, the intellectual property owner has had no choice but to resort to the courts if a simple request (or threat of litigation) didn't yield the desired result. This hardly seems unfair, since intellectual property disputes in all other areas of human interaction (product packaging, product names, print media, television, radio), even if international in scope, are resolved in the courts.

The sole exception to this rule is the Internet second-level domain name. Historical accidents, recent trends in commerce, and clumsy policymaking by NSI have made the Internet second-level domain name the most hotly contested asset on the Internet, and have led to a dispute-handling regime that harms the Internet by injecting extraordinary uncertainty into the business plans of law-abiding members of the Internet community.

What is it that makes second-level domain names so different from everything else on the Internet? Why would a government contractor that otherwise keeps the lowest possible profile choose to inject itself into the public debate by enacting such a controversial policy regarding second-level domain names? Two factors provide at least part of the answer to these questions.

Getting a particular domain name is viewed as crucial by many businesses. The Unum Corporation tried to obtain the domain name, learned that it had already been registered by someone else, and sued the owner to get it.(4) Explaining why they needed the domain name so urgently, Unum stated in court papers that:

[I]nformation on companies and their products and services is usually located on the Internet by typing in a domain name containing the company's name or trademark followed by ".com" (e.g. "".) As such, a company's ability to use a domain name on the Internet consisting of its company name followed by ".com" is important to its ability to successfully market, promote and sell its products and services.(5)

It is important to appreciate that even if these views are not entirely justified (at the time Unum's papers were filed with the Court it was easy, for example, to find all mentions of Unum on the Web through Alta Vista and other search engines), such views are nonetheless strongly held by many large companies.

Losing a domain name can mean going out of business. A domain name (note that for the balance of this chapter the term "domain name" will be used as a shorthand for "second-level domain name") is important because its loss would at the very least cause disruption and monetary expense, and in many cases would put a company out of business. Roadrunner Computer Systems, Inc. ("RCS"), an Internet service provider in New Mexico, had some 700 customers who relied on as part of their email addresses.(6) After NSI wrote to RCS stating that its domain name would be cut off in 30 days, RCS sued NSI for a court order blocking the cutoff. RCS's president stated under penalty of perjury that loss of the domain name "would be disastrous" and that one-fourth of the customers would be lost, in part because all of the customers would have had to change their email addresses.(7)

The triangle

Each trademark domain name dispute necessarily involves three parties -- the domain name owner, the trademark owner, and the registration authority. Each party has interests differing greatly from those of the others.

The domain name owner. For a domain name owner the predominant interest is predictability. The domain name owner doesn't want its domain name to be taken away precipitously any more than it wants to be evicted from its physical space or cut off from electrical power. For many Internet-related businesses, in fact, physical eviction or loss of electric power would be far more easily remedied than loss of the domain name. How can a domain name owner protect against loss of its domain name? Prior to NSI's July 1995 policy a domain name owner could protect against loss of its domain name by the simple step of avoiding infringing anybody's trademarks. Starting in July of 1995, however, that wasn't sufficient; one had to be prepared to sue NSI as well.

The trademark owner. For a trademark owner there are really two interest areas. The first arises if trademark infringement is going on, in which case the trademark owner would like to get the infringement stopped right away; a subsidiary concern is reducing the cost of getting it stopped. The second interest area arises solely as a consequence of the NSI policy, and involves the trademark owner that wishes it could have a particular domain name, and learns that the domain name has already been taken by someone else who isn't infringing the trademark owner's trademark. (The law of trademark dilution is discussed below and presents special problems on the Internet.) A trademark owner in this situation won't get anywhere in court, but the NSI policy provides an alternative mechanism for taking away such a domain name.

The domain name registration authority. The main interest of a domain name registration authority (of which there are several hundred around the world, one for each top-level domain) is getting its job done well. At present, about half a million domain names have been registered in the top-level domains administered by NSI; by comparison all of the other domain name registration authorities of the world combined probably account for only a few tens of thousands of domain names. This makes NSI's interests of particularly great concern, and NSI has stated many times in recent months that it is not only interested in getting its job done well, but is also interested in trying to avoid being sued.

Most of the domain name registration authorities of the world are volunteer organizations or are more or less affiliated with universities and government agencies, and thus have few interests other than those of their users, the domain name owners. NSI, in contrast, is engaged in numerous lines of business in addition to the administration of Internet domain names including the installation and maintenance of computer networks. NSI's corporate parent SAIC has revenues of US$1.9 billion, and has 20,000 employees and over 300 office locations around the world, and is a major contractor in information technology, systems integration, energy, environment, medical and health care systems and transportation.(8) It is thus possible that NSI has interests other than those which it has publicly disclosed. For example, NSI's and SAIC's customer base may have a greater overlap with the set of trademark owners than with the set of domain name owners.

In any event, the three parties to any dispute thus form a triangle, and it is important to consider the interests (disclosed and undisclosed) of all three parties when figuring out what to do about domain name trademark disputes.


The origins of the Internet lie in documents called RFC's, the very name of which ("requests for comment") communicated a consensus-based culture. RFC 1591, entitled Domain Name System Structure and Delegation, sets forth a simple role for the domain name registration authority:

In case of a dispute between domain name registrants as to the rights to a particular name, the registration authority shall have no role or responsibility other than to provide the contact information to both parties.

This remains, by definition, the default policy of any Internet domain name registration authority unless it establishes a different one. The domain name policies of the world's several hundred registration authorities have been surveyed and most of them follow RFC 1591 in whole or in part.(9)

Of course it is possible to imagine a trademark owner suing a registration authority in connection with a domain name dispute, but in such cases the registration authority need not worry much about legal bills or about financial liability. The reasons for this are simple. First, in most trademark cases (whether on the Internet or elsewhere) there is no award of money damages; instead, the only action taken by the court is to order someone to do something or to stop doing something. Second, if a trademark owner sues the registration authority it is generally done merely for procedural reasons, to ensure that at the end of the case the registration authority will comply with whatever outcome is ordered by the court. Third, at least for the registration authorities that have a history of abiding by court orders and for living up to their own commitments and policies, there is little reason to think that a court would impose any financial liability on a registration authority. Fourth, there are no known cases of a domain name registration authority being sued (by a trademark owner) for damages as distinguished from the authority being sued simply to do something with the domain name at the end of the case.

By about 1994 there had been a few highly publicized cases in which individuals had registered domain names (e.g., in ways that had angered trademark owners. Journalists quite cheerfully gave coverage to these mosquito-and-elephant stories in which large corporations had been slow to appreciate the Internet and found the domain names that they might have wanted, already taken by others. The impression given was that the Internet was the site of a sort of Oklahoma land-rush with enterprising individuals trying to stake out the likes of,, and and retire on the proceeds from the later sale of the domain names to their namesakes.

The com domain name was administered by SRI until April of 1993 when NSI took over that duty. In 1994 a company called Knowledgenet tried to obtain the domain name but found it was already taken. In December of 1994, Knowledgenet sued several parties in Chicago federal court, including the domain name owner and NSI.(10) From court docket entries, it is apparent that NSI (represented by the Shaw, Pittman firm in Washington) expended much lawyer time and money trying to get the case dismissed or transferred to Virginia. NSI never did succeed at getting the case dismissed or transferred, and settlement talks dragged on until the summer of 1995. The last docket entries in the case, in July and August of 1995, show that the Court was waiting for NSI and the plaintiff to file settlement papers, yet gives no indication that they were ever filed. Oddly, the court record shows the case as "terminated", yet nothing in the court record shows how the case ended, or if it ever did.

Meanwhile, in March of 1995 NSI was purchased by SAIC, a large but little-known defense contractor based in San Diego. SAIC's outside counsel, the Gray Cary firm in California, was given the task of drafting a domain name policy that would minimize the incentive for a trademark owner such as Knowledgenet to sue NSI in future. The result was the July 23, 1995 NSI domain name policy.(11) This policy created a tribunal paralleling the regular court system, but differing from the regular court system in that it was exceedingly sympathetic to and biased toward trademark owners. A trademark owner that wanted NSI to cut off someone's domain name had to do nothing more than write a letter to NSI stating that it owned a registered trademark identical to the domain name, and NSI would cut off the domain name after 30 days. (NSI would write what is now referred to in the Internet community as a "30-day letter" to the domain name owner.) The intention was apparently to promise ahead of time to do almost anything a trademark owner would have asked for in court, thus making it unlikely that the trademark owner would bother to sue NSI. The intention was also, apparently, to respond to the mosquito-and-elephant stories by providing a way that the elephants could get the domain names they had been slow to register. One of NSI's lawyers from the Gray Cary firm has said "I represent [NSI] and assisted in drafting the domain name dispute resolution policy and this type of problem is the reason for the policy. It permits the owners of registered trademarks to have special relief."(12) (emphasis added.) At the time the Gray Cary firm drafted the NSI Policy, it was listed as the legal representative on more than 1300 US trademarks and trademark applications for various clients.

This policy proved to be quite popular with trademark owners (presumably because of the "special relief" designed into it by its authors), who immediately began writing such letters to NSI. NSI initiated some ten cutoff proceedings the following month (August 1995) and the number of cutoff proceedings increased to a peak of about fifty in March of 1996. By May of 1996 some two hundred domain names had been cut off at the request of various trademark owners. In addition, the existence of the NSI policy resulted in many additional domain names changing hands privately since a trademark owner had merely to threaten to use the NSI administrative procedure and a domain name owner might give up the domain name (without the trademark owner actually having to write the letter to NSI) knowing the nearly inevitable result would be loss of the domain name anyway. In other words, the effects of the policy went far beyond the several hundred domain names that were the direct subjects of NSI cutoff proceedings.

From the trademark owner's perspective the NSI policy was a godsend. Instead of going to court with all the attendant drawbacks of doing so (the cost, the risk of a countersuit, and the risk of Rule 11 sanctions(13) or loss of the trademark if the court determined the case had been improperly brought), the trademark owner had merely to spend 32 cents and NSI would (1) assume the risk of countersuit and (2) cut off the domain name. Essentially NSI was offering (and still offers to this day) a free legal-services clinic for trademark owners, offering them "special relief".

What the NSI policy ignored (and continues to ignore to this day, despite two more policy revisions since the July 1995 policy) was that trademark owners were using its policy to win cases they could never have won in court. A trademark owner that had no bona fide case of trademark infringement against a domain name owner, but that was merely covetous of a domain name, could quite easily get it taken away from the domain name owner simply by writing the letter to NSI, or by threatening to do so.

Each 30-day letter which NSI mailed to a domain name owner (stating that the domain name would be cut off in 30 days at the request of a trademark owner) represented a risk of litigation for NSI, of course, since it would be possible a domain name owner would go to court to block NSI's threatened cutoff. But one can speculate that the drafter of the NSI policy looked to see who tended to have more money (owners of well-known trademarks), who was more likely to sue (owners of well-known trademarks), and which side NSI had leverage over (domain name owners) in selecting a corner of the triangle to favor. And indeed this assessment proved sound: during the first eight months that the policy was in force, not one domain name owner fought back in court. The trademark owner always won, and NSI avoided getting sued.

Between March and July of 1996, however, six different domain name owners (represented by six different law firms) who received 30-day letters had decided to fight back.(14) In each case the domain name owner sued NSI and scheduled a hearing at which the judge would be asked to issue a court order blocking the cutoff. Significantly, in five of the six cases, NSI contacted the domain name owner prior to the hearing date and agreed to scrap its policy so as to avoid being ordered to do so. (In the one case where NSI failed to do this, the case in Colorado, the judge signed an injunction ordering NSI not to cut off the domain name.)(15) It became clear that if a domain name owner that was not infringing anybody's trademarks received a 30-day letter from NSI, the most straightforward way of staving off loss of the domain name was to sue NSI.

Under NSI's second(16) and third(17) policies (in effect from July 1995 to September 1996) there was only one other means (besides suing NSI) by which a non-infringing domain name owner that had received a 30-day letter could keep from losing a domain name,(18) and that was to rush to a country in which a trademark registration could be obtained quickly. The reason is that under the NSI policy, the recipient of a 30-day letter was invited to supply proof of ownership of a trademark registration before the 30 days had expired, in which case NSI would not cut off the domain name. (The drafters of the NSI policy had presumably selected 30 days on the assumption, now known to be mistaken, that it was impossible to obtain a trademark registration in that short period of time.) Only one country, Tunisia, has been found that can provide a trademark registration (and the special "certified copy" of the registration which NSI demands) in such a short time, and several recipients of 30-day letters have used Tunisian trademarks to attempt to stave off loss of their domain names. NSI's fourth policy,(19) however, which is effective on September 9, 1996, eliminates this option by stating that the domain name owner can use a trademark registration to halt the NSI proceeding only if the registration was obtained prior to the start of the NSI challenge proceeding. This leaves suing NSI as the only reliable remaining line of defense for a non-infringing domain name owner that happens not to have its own trademark registration already.

Under NSI's fourth policy the only safe harbor for a domain name owner (short of suing NSI upon receipt of a 30-day letter) is to do whatever is required to obtain a trademark registration identical to the text of the domain name. Obtaining a US trademark takes a year or more, however. (Indeed even a domain name owner that somehow learned of NSI's poorly publicized policy change in July 1995 and immediately commenced applying for a US trademark would probably not have received it by now.) Thus the domain name owner that wishes to be safe against NSI 30-day letters starting right now has no choice but to obtain a Tunisian trademark registration. Of course, most domain name owners will not do this -- it seems silly to obtain a trademark in a country where one has no intention of doing business. Instead, it is fair to assume that most domain name owners will prepare and file trademark applications with the US Patent & Trademark Office (because most of the domain name owners in domains administered by NSI are in the US). As there are presently about half a million domain names administered by NSI, this can reasonably be expected to result in at least several tens of thousands of trademark applications filed with the US Patent & Trademark Office (USPTO) that would otherwise never have been filed. This will result in a severe overload of the existing trademark examining corps within the USPTO. (The USPTO has indicated that it plans to conduct a public policy-making proceeding in coming months to attempt to figure out what to do about this problem.)

Fundamentally the problem with the present (fourth) NSI policy is that one cannot protect oneself from loss of a domain name merely by avoiding infringing anybody's trademarks. (In all other areas of human endeavor, such as product packaging or naming, merely avoiding infringement does provide such protection; it is only in the specific area of NSI-administered domain names that this great risk presents itself.) One is forced to obtain a trademark registration oneself, a process that takes many months or years in most countries of the world. A related problem is that in the US, at least, trademark registrations are not available to all applicants as of right. To obtain the registration it is necessary to state under oath that one is actually using the trademark to indicate the origin of goods or services, and that this is taking place in interstate commerce. But there are probably a substantial number of domain name owners who do not, in fact, use their domain names to indicate the origin of goods or services, but merely use them in connection with a company name or line of business. And there are probably a substantial number of domain name owners who, even if they do use a domain name to indicate the origin of goods or services, do not do so in interstate commerce. All such domain name owners are stuck between a USPTO that will not give them a trademark registration, and NSI that says that nothing but a trademark registration provides a defense to a 30-day letter.

It should be noted that while the NSI policy does give "special relief" to trademark owners, it does not do everything that a trademark owner could want. Most importantly, from the trademark owner's point of view, the NSI policy falls short because winning a domain name challenge proceeding does not mean the trademark owner gets to have the domain name. Instead, what NSI does with the domain name is to place it "on hold". To get the domain name taken "off hold" requires getting a court order, or paying the domain name owner enough money to induce it to give up the domain name. Let's see how this works out for trademark owners in the context of the various scenarios that prompt trademark owners to initiate challenges.

For the trademark owner that merely wants to bring a halt to some infringing activity, the NSI policy is fully satisfactory. By getting a domain name placed "on hold", the trademark owner achieves its goal of stopping the activity, and does so at nominal cost (less than a dollar), while leaving to NSI much of the risk of getting sued by the domain name owner.

But many trademark owners want more than simply a domain name placed "on hold" -- they want the domain name itself. (As mentioned above, this can happen either because the domain name owner's actions actually infringe the trademark owner's rights, or simply because the trademark owner is covetous of the domain name.) For that, the trademark owner has no choice but to go to court, just as it would have if NSI had kept RFC 1591 as its policy. And the trademark owner might just as well sue not only the domain name owner but also NSI, as a way of making sure that NSI would comply with whatever court order might follow.(20) Indeed this very situation (a trademark owner suing a domain name owner and additionally naming NSI to be sure it will comply with court orders) has arisen four times since July of 1995.(21)

Anti-dilution laws and the Internet

Most trademarks are non-unique. There is a Yale lock company and a Yale University, and neither one will ever be able to block the other from using the word "Yale". As a general rule the owner of a trademark will only be able to block some other use of the mark if, in the opinion of the court, that other use gives rise to customer confusion. Because a court will not find customers of the lock company to be confused as to the origin of the University's services, and vice versa, trademark law will not allow either to block the other. A court may compare, for example, the goods and services offered by one company with the goods and services of the other, and if there is no overlap the court may find that there is little likelihood of confusion.

A tiny fraction (probably less than 1%) of trademarks are indeed unique. Kodak, Xerox, and Exxon are all coined names used solely by their namesakes. Legislatures have responded to the requests of the Kodaks and the Xeroxes of the world by enacting so-called anti-dilution laws which permit the owner of a unique trademark to block someone's actions regardless of whether or not the actions give rise to confusion. The accused infringer in a dilution case will find it no defense that its goods or services have no overlap with those of the trademark owner; the only defense is to show that the trademark is not the kind of mark that is protectable by the anti-dilution law, for example that it is not unique. For the drafter of an anti-dilution law the single most important task is to provide a cogent and workable criterion for determining whether a particular trademark deserves the special status of being "undilutable", namely, whether the court will enjoin accused infringers without the trademark owner's having to show that the actions cause marketplace confusion. Many states of the US have anti-dilution laws and the language of those laws often defines an undilutable mark as a coined mark (i.e. that it is not a word that was previously used) or as a unique mark (a mark that is used by only one company).

In January of 1996 the US Congress enacted a federal anti-dilution law.(22) In a move that virtually assured frequent collisions on the Internet for years to come, the law fails to provide any but the fuzziest language as to which trademarks deserve the special status of being federally undilutable. It merely says that such a trademark has to be "famous", and then provides a nonbinding list of eight factors which courts are free to apply or not as they see fit in their efforts to determine whether a trademark is "famous". Sen. Patrick Leahy said this about the Act:

It is my hope that this antidilution statute can help stem the use of deceptive Internet addresses taken by those who are choosing marks that are associated with the products and reputations of others.(23)

It would thus not be surprising if every trademark owner were henceforth to claim that its mark is "famous" and thus argue that it need not show, in court, that the accused infringer is actually causing confusion. And that is exactly what has happened. During the months that have passed since the federal anti-dilution law has been enacted, every single US lawsuit by a trademark owner relating to a domain name has asserted the federal anti-dilution law. And each such lawsuit has been accompanied by a brief quoting the statement by Sen. Leahy. Surely not all of the asserted trademarks are in fact "famous" (or unique, or coined) and yet the claim is uniformly made by the trademark owner for the simple reason that there is always some chance, however remote, that the court may find the mark "famous".

The US Congress, by passing an anti-dilution law with such a vague definition of "famous", has fanned the flames of Internet domain name problems. As a practical matter this problem will only fade after some years as the US courts interpret the term "famous" and eventually provide some sharper dividing line between the tiny handful of trademarks that are deserving of special "anti-dilution" treatment and the vast majority of other trademarks that are not. Until then one may predict with near-certainty that every US lawsuit asserting a trademark will invoke the federal anti-dilution law, which will exacerbate the problems with the present NSI policy. For example, under the present NSI policy a trademark owner that wishes to have NSI cut off a domain name must first write a letter to the domain name owner stating "unequivocally and with particularity that the registration and use of the Registrant's Domain Name violates the legal rights of the trademark owner."(24) Since the federal anti-dilution law is one of the "legal rights" which any trademark owner can assert (in the hopes that its trademark will be deemed famous), and since at present no one knows what is "famous", then any trademark owner can quite easily make the assertion required by NSI based on the federal anti-dilution law, even if the domain name owner is not causing any confusion.

Proposed solutions for the domain name trademark problem

It has been often stated, as if it sheds light on the matter, that the trademark system does not map onto the domain name system. There may be some truth to this, but if so the problem is neither new nor unique to domain names: there is likewise no workable mapping from trademarks to alphanumeric telephone numbers, to postal addresses, to names of pedigreed dogs or horses, or to stock exchange ticker symbols. (It often happens, for example, that a company making arrangements to be listed on a stock exchange finds that the ticker symbol it prefers has already been taken by some other company.)

One proposal is to establish new top-level domains corresponding to each of the several dozen international trademark classifications.(25) While the maker of York air conditioners and York peppermint patty candies cannot both have, on this proposal one might have york.mach (for machinery, say) and the other might have Critics of this proposal have pointed out, correctly, that even within a single international trademark classification it is commonplace to find dozens of companies with the same name. They coexist (in the real world) because their lines of business do not in fact overlap even though they happen to be in the same trademark classification. Yet they would lead to collisions even if this proposal were implemented. Critics of this proposal also point out, correctly, that trademark owners tend to want to protect not only their present line of business but also all possible future lines of business. Thus, if the trademark-classification domain names were announced, the Disney company (for example) would presumably instantly sign up for and disney.mach simply to protect possible future brand name extensions.

It has also been stated that from its beginnings the domain name system was never intended to be a directory system. It is slightly disingenuous to say this, since from the beginnings of the domain name system nobody ever expected, for example, that would map to anything other than MIT, or that would map to anything but Harvard University. Likewise it is fair to say that once the com domain was defined, nobody would have expected to map to anything but the Xerox Corporation. Indeed from the beginnings of the domain name system there was a not-very-clearly-articulated hope that domain names would desirably be "guessable" to the extent possible, just as Unix commands are "guessable" (mv means "move", cp means "copy", etc.) And until Web search engines became common in about the summer of 1995, there were only about four ways to find out somebody's domain name (and, for example, their email address): (1) wander around in Gopherspace from site to site hoping to find the answer; (2) use Whois to search for the organization name; (3) guess that the domain name might be the organization name or some variant of it, or (4) call them up on the telephone and ask.

These four ways of finding a domain name (and/or email address) lead to a perception by large corporations that the only acceptable domain name is the corporation name followed by com. (Recall the statement by the Unum Corporation to this effect, quoted above.)

Many people have remarked upon the great speed with which the World Wide Web transformed the ways people use the Internet, and probably very few of us imagined, even as recently as five years ago, that a hypertextual environment with free access to a meaningful fraction of the sum of human knowledge would become a fixed part of society in 1996. But even after the force of the Web became clear a couple of years ago, few would have predicted the almost instantaneous development of search engines (e.g. Lycos, Infoseek, Alta Vista) that would offer extraordinary searching power and convenience for free to everyone.

The fact is that in 1996 one can find the domain name for any large corporation quite easily even if it happens not to be the corporate name followed by com. It is a trivial matter to plug in the company name with any search engine and very quickly to find the company's web site. One may then make a bookmark in one's web browser, and thus never need to type in the domain name even once.

It is likely that new meta-levels (in addition to the search engine capabilities) will be developed and imposed between the user and the domain name system, making it less and less important for a company to have its exact company name as a domain name. Such developments will probably be as difficult to foresee as the search engines were.

Still other commentators have suggested that the way to lessen or eliminate the domain name trademark disputes is the addition of new top-level domains (TLDs) which would compete with com and would be administered by registration authorities other than NSI. Frequently mentioned proposals for new TLDs include alt, biz, and corp. It has been suggested that the availability of such TLDs would "relieve the pressure" on com by providing other places where companies could obtain domain names identical to their company names. But there are several reasons to predict that adding such new TLDs won't help this problem.

First, if you have spent years developing a business that relies upon some particular com domain name, it is no comfort whatsoever to be told that if you want to, you can register with some other TLD. To do so would lose all the accumulated goodwill and web browser bookmarks that have assured a future for the business.

Second, there are already over a hundred top-level domains (including, for example, over a hundred two-letter domains corresponding to countries of the world). Many hundreds of domain name challenges (and many lawsuits) have been brought by trademark owners who could just as well have registered in any of those hundred-odd domain names, but what they wanted was a com domain. Nothing else was good enough for them. There is little reason to think that adding more top-level domains will change this perception that com is the place to be.

Third, it is simply wrong to say that the present domain name trademark problems stem from "pressure" on the present com domain; it is simply wrong to say that com is full or that the present addressing space isn't big enough to accommodate all who wish to have distinct domain names. To see this, consider that any proposed additions or changes to what would be to the right of the "dot", that supposedly would cure some problem, could just as well be added or changed to the left of the "dot". Suppose that is taken already, and that United Air Lines wants a domain name. While some proponents of additional TLDs would have is believe the best (or only) way to accommodate the airline is to create a new TLD air (to permit the airline to have a domain name united.air), the reality is that the fiddling could happen just as well to the left of the dot, yielding perhaps or or There is plenty of address space in com and it will never run out.

This is not to say that it would be a waste of time to set up new TLDs. It would probably be quite desirable to have a number of new non-geographic TLDs, distinguished not so much by the connotations of the letters making up the domain, but rather by differing policies or levels of service of the registration authorities. One registration authority might distinguish itself on price, charging little in the way of annual fees. Another might distinguish itself by promising not to cut off a domain name unless ordered to do so by a court. Yet another might promise to have efficient and accurate billing and invoicing.

Perhaps with the passage of years the perception that it is crucially important to have an easily guessed com domain will fade. This may happen because of a more widespread appreciation for search engines, or by shifting fashionableness of particular top-level domains, or because of some future meta-level change (e.g. a shift away from character-based input to speech recognition) in the way people interact with the Web and the Internet. But for the near future there are half a million stakeholders, present-day domain name owners who need predictability and stability to be able to justify further investment of money and sweat and human creativity into their Internet-related businesses. The present regime in which a domain name can be taken away on only 30 days' notice from someone who isn't doing anything wrong, has to change. That's one reason that meaningful debate needs to take place toward development of good domain name trademark policies. NSI, the holder of a position of public trust regarding nearly all of the domain names on the Internet, can't be permitted to continue to develop its policies in secret with no meaningful stakeholder involvement.

A second reason why such debate is important is that it can help the hundreds of present domain name registration authorities as they continue in the fulfillment of their duties, and as they consider whether to make any changes to their policies.

A third reason why such debate is important is that it can help the founders of newly created top-level domains as they draft their policies.

Designing a good domain name trademark policy

This chapter now turns to the question of how one can design a better domain name trademark policy. It is fashionable nowadays to say that the pace of technological change is so great that the legal system can't and won't be able to keep up. This view is offered in support of any number of propositions, such as that it should not be copyright infringement to copy what is found on the Internet, that the patent system should not be applied to the works of computer programmers, and that we should simply declare that the world of domain names has no connection to the world of trademarks. Leaving those debates aside, this chapter reflects the author's view that whether we like it or not, the legal system is in fact the way that disputes in modern society get decided if and when the other mechanisms (asking nicely, negotiating, mediation, NSI challenge proceedings, etc.) don't work. Thus we have little choice but to consider how a particular dispute (or class of disputes) would be decided in a court of law, as we consider how to design a policy. When a policy yields outcomes that differ from what a court would do, the policymaker can be sure of getting sued. When the policy yields the same outcome, the policymaker is much less likely to get sued.

This simple observation helps to show how wisely RFC 1591 was drafted. Under RFC 1591, a registration authority doesn't take action. As a result, it doesn't take action that would differ from what a court would do. Implicit in RFC 1591 is that the registration authority will obey the order of the court. This means that by definition the registration authority does exactly what a court would do (because it does what the court tells it to do). Under RFC 1591, nobody is going to have much of a reason to sue a registration authority (other than for the mere procedural purpose of making sure that it is a party to the case and thus can be relied upon to obey the court order).

In every case to date in which a trademark owner sued a registration authority, the trademark owner did not merely sue the registration authority but also sued the domain name owner. It would, however, be unrealistic to suppose that the former would never happen. In other words, it is possible to imagine a trademark owner suing the registration authority alone, perhaps seeking a court order directing that the domain name be given to the trademark owner. In such a case, how would the registration authority deal with the lawsuit? How could it bring the lawsuit to an end? Where would the money come from to deal with months or years of active litigation? The answer, it turns out, can be found in a corner of the legal system that most lawyers hear about for part of one day in their first year of law school, and promptly forget: the doctrine of interpleader.

The law of interpleader had its origins in banking law. If two parties (would-be heirs of a deceased accountholder, for example) both wanted the contents of a bank account, one party might sue the bank to try to get the money. Under the doctrine of interpleader, the bank does not have to expend any money or lawyer time defending the lawsuit. The bank instead simply makes a formal tender of the asset to the court (the asset itself normally does not physically change hands) and advises the court that it will dispose of the asset in accordance with the court's orders. The other party (the one that did not sue the bank) is served with papers inviting it to participate in the interpleader action. The parties present their cases, the court decides who gets the money, and that is the end of the matter.

For the bank, the interpleader procedure cuts its legal costs to a minimum. Upon getting sued, the bank presses a button on its word processor and prints the interpleader papers. From that point onwards the bank's lawyers have no duty other then to watch the incoming mail for a court order informing the bank that the case is over and that the money should be paid out in some particular way.

Interpleader works just as well in the case where one of the parties sues not only the bank but also the other party. In that case, too, the bank merely interpleads the asset and waits for the dust to settle and for the court order to arrive.

A domain name registration authority can do the same. If sued, it simply tenders the asset (the domain name) to the court, stating that it will comply with the final order of the court. In this way it minimizes legal expense. A domain name registration authority that behaves in this way consistently for a period of months and years will eventually find that trademark owners won't bother to name it as a defendant since they will trust it to obey the court order anyway.

It is intriguing (but futile) to wonder what would have happened if NSI had responded to the Knowledgenet lawsuit by means of interpleader rather than the costly jurisdiction and venue battle. NSI's legal fees would presumably have been far smaller, and perhaps it would not have felt the need to enact what was apparently a hastily contrived policy in response to those fees.

Which court? The preceding discussion ignores that the Internet has no boundaries, and that the three corners of the triangle may be in three different countries. It refers to "the court" as if there were only one. But it is necessary to consider what can happen if several countries are involved. Suppose the registration authority is in country A, the domain name owner is in country B, and the trademark owner is in country C.

From the trademark owner's point of view it would, of course, be desirable if the lawsuit could be in its country C, but there are at least two reasons why C is likely to be unsuitable as a forum -- it might easily not have jurisdiction over the domain name owner, and the court might not be capable of having its orders enforced outside of its own country. Alternatively, the trademark owner could sue the domain name owner in its country seeking an order directing the domain name owner to transfer the domain name. This is what the trademark owner would have to do in the case of most other types of disputes, after all. Finally, the trademark owner could simply sue the registration authority in its country, and wait for the registration authority to interplead the domain name and serve its interpleader papers upon the domain name owner. The effect of the proposed domain name policy doesn't harm the trademark owner because its only effect is to add to the number of places where the trademark owner could file its lawsuit (two countries in total), in comparison with the list of places where the trademark owner could sue if the dispute were in some other area (e.g. a fight over a third-level domain name), namely the country where the domain name owner is located.

From the domain name owner's point of view the proposed policy might seem unfair. After all, the domain name owner will be receiving interpleader papers inviting it to defend itself in some other country (the country where the registration authority is located). The expense and disruption of traveling to that country to defend itself could be great. This isn't much of a problem in the simple case where a top-level domain is newly created and the registration authority has disclosed the policy from the outset. In such a case the domain name owner will presumably have taken the policy into account when choosing which top-level domain to register in. If the domain name owner didn't like the notion of having to travel to country A to defend itself, or didn't trust the courts of country A to decide cases fairly, then the domain name owner would presumably have chosen a different TLD in the first place.

The more complicated case is one where a registration authority announces for the first time that it plans to use interpleader, in other words, in which the domain name owner had not previously been told this. It might be suggested that this is unfair. One must take into account, however, that if a domain name owner selects a registration authority located in a foreign country, it cannot claim to be surprised if finds itself having to deal with the courts of that country. What's more, the law of interpleader is old, and in particular is older than the Internet. No one can claim to be surprised if some well-established legal procedure gets followed in a case where the procedure was and is applicable.

NSI's recent implementation of its third policy uses a backwards variant of interpleader. In each of two recent cases,(26) NSI wrote to a domain name owner stating that its domain name would be cut off in 30 days, after which the domain name owner sued NSI asking for a court order that the domain name not be cut off. NSI's response was to file a brand-new lawsuit initiating an interpleader action against both the domain name owner and the trademark owner. In a case where the domain name owner was not infringing anybody's trademarks, this is doubly unfair, as it forces the domain name owner to incur the expense of defending itself in not one but two actions. (In one of the cases the domain name owner is having to defend itself not only in two actions but in two different courts in two different cities.)

Third-level domains should be emphasized. From the beginnings of the domain name system, it was contemplated and desired that third-, fourth-, and higher-level domain names would be commonly used. For example, the Kennedy School of Government uses a third-level domain ( rather than a second-level domain (e.g. for its web site. This is a very good thing for the Internet for several reasons. First, it saves some workload for the root-level servers because they need only answer lookup requests for harvard and not for ksg. Instead, servers operated by Harvard answer lookup requests for ksgwww. This spreading-out of the DNS workload was intended by the designers of the Internet. Second, it saves work for the administrator of the edu domain, because the creation and deletion of third-level domains such as ksgwww can be done locally by the administrator of the domain. Finally (and, most importantly for Internet policymakers) the use of third- and higher-level domains expands the domain name address space and reduces pressure on the root-level domains.

Some of the most highly visible domain name disputes have involved companies that already have one or more Internet domain names (for example, the company name followed by com) and that wish, in addition, to possess domain names based on the names of products made by the company. Hasbro, a maker of children's games, has had the domain name since 1994. Hasbro's products include a game called Clue. If Hasbro wanted to set up a web site, say, for its game Clue, the most Internet-friendly way to do this would be to program its domain name servers to create

Instead, Hasbro went to NSI and attempted to register, only to find that a company called Clue Computing in Colorado had beaten them to it. Were it not for NSI's policy, Hasbro would have had no choice but to accept that someone else owned and that perhaps was the way to set up a domain name for the board game. Stated in different terms, if Hasbro had simply filed a lawsuit to attempt to obtain, Hasbro would have had the case dismissed and might have faced sanctions. There are hundreds of businesses called "Clue" and none of them has legal grounds for taking from another.

NSI's policy, however, provided another avenue for obtaining the domain name. Hasbro simply wrote to NSI stating that it owned a trademark registration for "Clue", and NSI sent a 30-day letter to Clue Computing. Clue Computing now faces not only the expense of the lawsuit it brought against NSI to block the cutoff of the domain name, but also the expense of an additional lawsuit which NSI filed against it and against Hasbro in response. These are expenses that Clue Computing would never have faced if NSI had retained RFC 1591 as its policy.

Another example of the pursuit of second-level domains when third-level domains use Internet resources more efficiently is a well-known automobile maker that is trying to take away from others the second-level domain names corresponding to its car models.

In its early days of domain name administration, NSI urged domain name owners to try to get by with only one or two domain names each. Now that NSI is collecting annual fees it has stopped making such recommendations. The Internet community and policymakers should find ways to exert pressure or at least moral suasion upon those who try to amass second-level domains when their needs could better be served by a single second-level domain and a number of third-level domains. Domain name registration authorities should encourage their customers to consider the use of a third-level domain rather than a second-level domain when possible. Lawyers advising domain name owners and trademark owners should try to help their clients understand the purpose of third-level domains.

This chapter refers repeatedly to conflicts over second-level domains, and recommends among other things that members of the Internet community that already have a second-level domain be urged and/or shamed into using third-level domains (e.g. The terminology is driven by the historical accident that nearly all of the highly publicized conflicts have been in the COM domain and have involved second-level domains such as But in many countries, the two-letter top-level domain (e.g. au for Australia) is subdivided into content-related second-level domains such as for educational institutions in Australia, for commercial entities in Australia, etc. The recommendations of this chapter apply mutatis mutandis. For example, in Australia if there are conflicts they are likely to be over third-level domains (e.g. in which case what should be encouraged is the use of fourth-level domains (e.g.

Policies should be diligently communicated to the stakeholders. One of the standard practices of the registration authority is that any and all policy changes should be diligently communicated to the stakeholders. While this might seem onerous at first blush, the fact is that the registration authority can use email for this purpose at little or no cost. The registration authority already has contact information for each domain name owner including email addresses. Failure to communicate this policy leads to profoundly unfair results. For example, under the NSI policy the trademark owner can ambush the domain name owner. The trademark owner will have been able to take as much time as desired to prepare for the eventualities of a domain name challenge. In contrast, the author of this chapter has spoken with many recipients of NSI 30-day letters, and almost without exception the domain name owner reports that the first time it ever knew of the NSI domain name policy was when NSI sent the 30-day letter (which has a copy of the policy attached). Many recipients of 30-day letters are domain name owners that registered their domain names long before July 1995 and thus could not have known, at the time they registered, of the policy. Had NSI done a trivially simple broadcast emailing to all of its domain name owner customers in July of 1995, then some of them would have filed trademark applications at that time with the USPTO, and might now have trademark registrations and thus be able to defend themselves against 30-day letters. [Note: Shortly after this paper was presented, NSI commenced a broadcast of email messages to the administrative contacts for all the domain names for which NSI is the registration authority.]

Policies should not be changed retroactively. One of the most controversial aspects of NSI's policy is that it makes a drastic change to the rules and purports to apply the change to domain names that were registered long ago. A domain name owner who registered a com domain in, say, 1994 did so at a time when the only way a domain name could be taken away was by court action, and a court would only take away a domain name if the domain name owner had done something wrong. The domain name owner presumably chose to invest time and money in its business with that in mind. Then came the July 1995 policy, and suddenly a domain name could be cut off even if the domain name owner wasn't doing anything wrong.

If a registration authority chooses to change a policy retroactively, extreme care should be taken in the design of the policy to avoid causing harm to those who registered domain names earlier and in good faith and who aren't doing anything wrong.

The registration authority should conduct its deliberations on an open record. Society benefits if disputes don't have to go to court. Society benefits if disputes get settled by the parties prior to any proceedings, in court or elsewhere (e.g. in some proceeding before the registration authority). Ordinary courts conduct their proceedings on an open record. Those who wonder if the court is fair can look at the record and reassure themselves on this point.

For the same reasons, the registration authority should keep an open record. This permits domain name owners and trademark owners to see that disputes are decided fairly. It likewise permits domain name owners and trademark owners to reach their own resolutions without having to involve the registration authority or the court, because they can simply predict for themselves what the outcome would be, and settle on those terms. Keeping an open record is particularly important if the registration authority has interests that are not fully disclosed and may conflict with the interests of its domain name owners. Indeed any proposed new registration authority should be strongly encouraged to reveal its other interests.

NSI conducts its proceedings in secret, and its past and present policies contain vague areas (e.g. what counts as "identical" as between a trademark and a domain name, and how strongly worded a trademark owner's letter must be to trigger an NSI domain name cutoff proceeding). This leaves the public with no way of knowing whether NSI is fair in its decisionmaking, and leaves disputants unlikely to be able to settle disputes because each may have a differing guess as to how NSI would decide the dispute.

Ideally a registration authority would leave it to the courts to decide disputes about the domain names that it administers, which would leave very little for the registration authority (since it would make few if any decisions) to disclose on its open record.

Immunities aren't needed

It has been suggested from time to time that what is needed is a legislative immunity from suit for domain name registration authorities. Their position is special, so the argument goes, because they hold a position of trust with respect to this extraordinarily important part of our modern society, the Internet. This chapter argues otherwise, and for several reasons.

First, there are many ways for a domain name registration authority to get sued other than by some trademark owner that has a gripe about the actions of a domain name owner (or is merely covetous of the domain name). A domain name registration authority might be sued by a visitor who slips and falls in the reception area, or by a creditor who claims a bill has gone unpaid. Such lawsuits are a simple fact of doing business with the public, and one must simply get used to it. There is no compelling reason why any particular category of lawsuit should be specially blocked by some legislative immunity.

Second, giving a domain name registration authority immunity from suit essentially grants a blank check to the registration authority to engage in arbitrary and capricious conduct without fear of having to answer for its actions. Domain name owners who have viewed NSI's past conduct with alarm are aghast at the prospect of having the court system taken away from them as a way of protecting themselves from NSI.

Third, as described above, the specter of liability arising out of the conduct of a domain name owner, cited by NSI as a justification for its controversial policy, is greatly exaggerated. No domain name registration authority has ever been held liable for the conduct of a domain name owner, nor has any trademark owner ever sued a domain name registration authority asking for an award of money damages from the registration authority. Uniformly the trademark lawsuits in analogous situations (for example, suing a publisher for a trademark infringement by an advertiser) have sought no more relief from the publisher than simply an injunction against future conduct, not an award of money damages. Indeed awards of money damages in trademark cases are quite rare, and in most trademark cases if any remedy is awarded at all it is merely an injunctive remedy.

Fourth, to the limited extent that a domain name registration authority has any legitimate concern about getting sued by a trademark owner, the registration authority can put a cap on its legal expenses by simply tendering the domain name to the court in an interpleader action. What follows is a period of little or no legal expense for the registration authority.


This chapter suggests that domain name registration authorities should simply adopt a first-come first-served policy for registration of domain names. Alternatively the registration authority could impose a simple, objective first-level screen by asking that the applicant show organizational papers (articles of incorporation, doing-business-as statement, birth certificate) indicating a putative right to use the proposed domain name, thereafter granting the registration on a first-come first-served basis. The policy should further provide that if someone sues the registration authority over a domain name (and if the suit is brought in a competent court in the jurisdiction in which the registration authority is physically located) then the registration authority will simply agree to obey the court's order. Interpleader should be invoked to bring before the Court the parties to the dispute. The registration authority should not presume to grant preliminary relief, deciding by itself when to cut off a domain name; such grants of preliminary relief should come only from competent courts. To the extent the registration authority does make any substantive decisions about the grant or cutting-off of domain names (and this chapter suggests that the registration authority should avoid doing so) the decisions should be duly memorialized on an open record. Policy changes should not, generally speaking, be retroactive in application unless there is no other way to proceed, and then only if due provision is made for the rights and interests of stakeholders. Policy changes should be communicated most diligently to affected domain name owners, by direct email for example.


(1) Partner, Oppedahl & Larson. Email:, web:

(2) I thank Bruce Albrecht <> and Dale Worley <> for pointing out this historic reference.

(3) Agmon, Jonathan et al. "What's In A Name?", <>, Shaw, Robert, "Internet Domain Names: Whose Domain Is This?", <>

(4) Unum Corp. v. Sanfilippo et al., N.D. Cal., 96-civ-1369. [Information at] The author was counsel for the domain name owner in that litigation, now concluded on confidential terms.

(5) Motion for Preliminary Injunction, Unum Corp. v. Sanfilippo et al., N.D. Cal. 96-civ-1369, <>.

(6) The author was counsel for RCS in that litigation, now concluded. [Information at] RCS got to keep its domain name.

(7) Declaration of Jane Hill, Roadrunner Computer v. NSI, E.D. Va. 96-civ-413A, <>.

(8) <>.

(9) Gussis, Geoffrey, Global Top-Level Domain Dispute Resolution Policies, <>.

(10) Knowledgenet, Inc. v. Boone et al., N.D. Ill, 94-civ-7195, docket record at <>, complaint at <>.

(11) NSI Domain Dispute Resolution Policy Statement, July 23, 1995, <>.

(12) Radcliffe, Mark, partner in the Gray Cary firm, posting to Net-Lawyers discussion group, December 17, 1995.

(13) Rule 11 of the Federal Rules of Civil Procedure provides sanctions under certain circumstances if a lawsuit is brought, or some other court paper filed, without reasonable inquiry having been made as to the basis for the lawsuit or paper.

(14) The cases are Roadrunner v. NSI, E.D. Va. 96-civ-413-A, DCI v. NSI, M.D. Tenn. 96-civ-429, Giacalone v. NSI, N.D. Cal. 96-civ-20434, Clue v. NSI, District Court, Boulder County, Colorado 96-civ-694-5, DISC v. NSI, D. Colo. 96-civ-1551, and Regis v. NSI, N.D. Cal. 96-civ-20551. As mentioned above, the author was counsel in the Roadrunner case. Much information about these six cases is available at <>.

(15) Clue Computing v. NSI, District Court, Boulder County, Colorado, Case No. 96 CV 694, Division 5 <>.

(16) <>.

(17) <>.

(18) There is one other fact pattern which can be imagined, namely the case in which a domain name owner that happens to already have a trademark registration finds itself the recipient of a 30-day letter. In real life this doesn't happen, because the would-be challenger does a trademark search first to see if the domain name owner has a trademark registration, and does not bother to initiate the challenge if a trademark registration owned by the domain name owner is found in the search.

(19) <>.

(20) It might be thought that the trademark owner could dispense with naming NSI as a party because NSI's present policy says it will obey court orders. But NSI could change its policy in this regard at any time; recall that NSI is now on its fourth policy in thirteen months and has repeatedly said that it is entitled to change its policy at will on a mere 30 days' notice. It is a very trusting litigator indeed that would choose not to name NSI as a party when suing a domain name owner, given NSI's unpredictable behavior.

(21) The cases are Porsche Cars North America, et al. v. Chen, et al., E.D. Va. 96-civ-1006 (; American Commercial, et al. v. Sports & Leisure, et al., C.D. Cal. 96-civ-713 (; Panavision Int'l. v. Toeppen, et al., C.D. Cal. 96-civ-3284 (; and Prestone Prods v. Maynerd Collision, et al., E.D. Va. 96-civ-234 ( Much information about these six cases is available at <>.

(22) Federal Trademark Dilution Act of 1995, 15 USC § 1125(c).

(23) Cong. Rec. Dec. 29, 1995, S19312.

(24) <>, para. 5(b).

(25) Collier-Brown, David, On Experimental Top Level Domains <>.

(26) and

Return to the main Flawed Domain Name Policy page.

Last revised September 13, 1996.

This document is