Tim Berners-Lee
Creation Date: 2004-04, last change: $Id: TLD.html,v 1.23 2004/04/30 20:55:53 ijacobs Exp $
Status: TAG supported. See below.

Up to Design Issues

New Top Level Domains .mobi and .xxx Considered Harmful

In 2004 there were proposals to create new top-level domains which included .mobi and .xxx. There are major problems with these proposals. There are costs in general to creating any new top level domain. There are specific ways in which the ".mobi" breaks the Web architecture of links, and attacks the universality of the Web.

At their 14 May 2004 face-to-face meeting, the W3C Technical Architecture Group resolved to support this document, with Norman Walsh abstaining, and Paul Cotton recusing himself.


When the Internet was being collaboratively developed by a substantially technical community around a growing but still manageable Internet Engineering Task Force, the Domain Name System (DNS) evolved as a hierarchical solution to the problem of keeping track of which computers had which Internet Protocol (IP) addresses. The tree structure was an improvement over the previous flat space of host names. It reduced the chaos, by allowing new names to be allocated in sub-domains without recourse to a central registration system. Because the frequency of allocation of new names decreased as one ascended the tree toward the root, the actual cost was kept manageable.

As email and World Wide Web (WWW) use blossomed and became increasingly important, domain names crept out of the messages syntax for Internet protocols and crept into daily parlance. It then became valuable to own a short domain name. This turned domain name space into a limited commodity. After some tussles for control (ongoing at the time of writing) and some large amounts of money changing hands in some cases, the system has now settled down to a market-based one in which names can be rented, transfer value can be asked by the old owner of the new owner, and one-time and annual fees are typically payable by any domain to any company managing the higher domain. An anomaly was that unclaimed names were deemed to have no owner and no value, and were allocated in a "first come first served" frenzy in which speculators made great profits and held to ransom those who may have been considered the more logical owner of a name. This anomaly created great instability. It has costs, in that any trademark owner had to beware of parties who would register domains which included their trademark. The Example Manufacturing Company had to ensure that it owned not only example.com which it had used for email and Web site for many years, but also example.net and example.org to avoid unscrupulous competition setting up sites to benefit from Example's excellent reputation. As the business grew, Example had to also acquire example.fr and example.co.uk to ensure that confusion was minimized.

The fact was that the public memory was not for the domain name, but for the brand name which was sandwiched between www and .com. To this extent, in the world of memorable domain names, the hierarchicalization of the domain system had failed to happen. In the public's memory, example was the mark, and the difference between example.com and example.net merely a source of confusion.

As each node in the tree represents a potentially valuable asset, control of any subset of the tree is valuable. Control of the entire tree is managed by ICANN, which is set up to be a non-profit international institution, with the intent that it should as such carry the trust of the entire community in its efforts to manage the system for the common good. Control of subtrees such as .net, .com and .org is delegated to set of parallel registries whose business model is nominally the charging of registration and annual fees. There have been temptations for the registry companies to consider themselves owners of unclaimed names. Rumors have abounded about systems which would automatically rent a domain name about which a potential renter was inquiring, or would redirect traffic from an unclaimed Web site to their own Web site, and so on.

The Cost of Change

The top level of the domain name system, and to a lesser extent the IP address space, are the single weak, centralised, points of an otherwise strong, decentralised system. The Internet is a net, and the WWW is a Web, but WWW and email use DNS which is a tree, which has a single root. Although there are many benefits to a system with global identifiers, there are also costs, such as a single common DNS tree. As a community we have all decided that the benefits of the system (such as being able to quote example.com anywhere in the world and have it mean the same thing) outweigh the costs of the social systems required to ensure fairness in its operation. There is, however, great stress. ICANN is under constant pressure to alter its balance of power or modus operandi. It balances technical, academic, commercial, and governmental inputs. The whole issue of domain names has created a vast amount of concern. And because the DNS tree is so fundamental to the Internet applications which build on top of it, any uncertainty about the future creates immediately instability and harm.

Our first instincts, then should be not to change the system with anything but incremental and carefully thought-out changes. The addition of new top-levels domains is a very disturbing influence. It carries great cost. It should only be undertaken when there is a very clear benefit to the new domain. In the case of the proposed .mobi domain, the change is actually detrimental.

The Economics of Domain Names

In practice, for most domain name owners, the part between the "www" and the top level domain is their brand, or their name. It is something they need to protect. This means that in practice, a serious organization to avoid confusion has to own its domain in every non-geographical top level domain. For a large company, the cost of this may be insignificant. For a small enterprise, a non-profit organization or a family, the cost becomes very significant.

The chief effect of the introduction of the .biz and .info domains appears to have been a cash influx for the domain name registries. Example Inc. as mentioned above owns example.com, org and .net. Does it also have to buy .biz, .info, and .name to avoid confusion and the misappropriation of my name by others? Will I have to also rent "example.mobi" in case it want to make information available for people who use wireless equipment?

The market for second-level domains is a market for a limited resource. After an unstable period when the first come first served system was in play and greedy squatters grabbed domains simply for speculation, it has now settled down. Introducing new TLDs has two effects.

The first effect is a little like printing more money. The value of one's original registration drops. At the same time, the cost of protecting one's brand goes up (from the cost of three domains to four, five, ...).

The value of each domain name such as example.com also drops because of brand dilution and public confusion. Even though most people largely ignore the last segment of the name, when it is actually used to distinguish between different owners, this increases the mental effort required to remember which company has which top level domain. This makes the whole name space less usable.

Is it fair to reduce the value of these domains which have been acquired at great cost by their owners?

The second effect is that instability is brought on. There is a flurry of activity to reserve domain names, a rush one cannot afford to miss in order to protect one's brand. There is a rash of attempts to steal well-known or valuable domains. The whole process involves a lot of administration, a lot of cost per month, a lot of business for those involved in the domain name business itself, and a negative value to the community.


As we have seen, the choice of a tree structure for domain names is one which has costs and benefits, and the community currently accepts both. The cost of confusion, and of extra name registrations, is high. When the benefits of the new domain itself are small or negative (as we discuss below), then one looks for incentive. The large amount of money that has changed hands for domain names might lead a person to suspect that this was the motivation. Under these circumstances, to increase public trust, proposals from non-profit organizations would raise less suspicion.

The root of the domain name system is a single public resource, by design. Its control must be for and, indirectly, by the people as a whole. To give away a large chunk of this to a private group would be simply a betrayal of the public trust put in ICANN.

Specific Problems with .mobi

The different domains are introduced for different reasons, so we must answer this for each one. The ICANN list of proposals gives pointers to the proposals.

The .mobi domain is described as being for the use of a community. To quote the proposal, the target community for the .mobi TLD is:

This is in fact a mixture of reasons. It sounds as though there is a use for ".mobi" when the provider of a service intends it to be for the benefit of mobile users. There appears to be a desire to limit the use of ".mobi" to companies -- perhaps those in the group.

This domain will have a drastically detrimental effect on the Web. By partitioning the HTTP information space into parts designed for access from mobile access and parts designed (presumably) not for such access, an essential property of the Web is destroyed.

Device Independence.

The Web is designed as a universal space. Its universality is its most important facet. I spend many hours giving talks just to emphasize this point. The success of the Web stems from its universality as do most of the architectural constraints.

The Web must operate independently of the hardware, software or network used to access it, of the perceived quality or appropriateness of the information on it, and of the culture, and language, and physical capabilities of those who access it [WTW]. Hardware and network independence in particular have been crucial to the growth of the Web. In the past, network independence has been assured largely by the Internet architecture. The Internet connects all devices without regard to the type or size or band of device, nor with regard to the wireless or wired or optical infrastructure used. This is its great strength. From its inception, the Web built upon this architecture and introduced device independence at the user interface level. By separating the information content from its presentation (as is possible by mixing HTML with CSS, XML with XSL and CSS, etc.) the Web allows the same information to be viewed from computers with all sorts of screen sizes, color depths, and so on. Many of the original Web terminals were character-oriented, and now visually impaired users use text-oriented interfaces to the same information.

For a time, many Web site designers did not see the necessity for such device independence, and indicated that their site was "best viewed using screen set to 800x600". Those Web sites now look terrible on a phone or, for that matter, on a much larger screen. By contrast, many Web sites which use style sheets appropriately can look very good on a very wide range of screen sizes.

It is true that to to optimize the use of any device, an awareness on the part of the server allows it to customize the content and the whole layout of a site. However, the domain name is perhaps the worst possible way of communicating information about the device. Devices vary in many ways, including:

and so on. While with the current technology, the phrase "Mobile" may equate roughly in many minds to "something like a cell phone", it is naive -- and pessimistic -- to imagine that this one style of device will be the combination that will endure for any length of time. Just as concepts such as the "Network PC" and the "Multimedia PC" which defined profiles of device capability were swept away in the onrush of technology, so will an attempt to divide devices, users and content into two groups. Small devices will have high bandwidth. Devices with large screens will sometimes have small bandwidth. Some "mobile" phones will be permanently mounted on kitchen walls. The range of digital assistants will continue to evolve.

There are good ways to deal with and derive the greatest benefit from the growing diversity of client devices. The adaptation may occur on the client side, the server side, or both. For example, the CC/PP specifications provide a framework for a client device to describe its capabilities in great detail to a server. This is based partly on the UAPROF (User Agent Profile) specifications developed by the mobile phone industry. Also, the HTTP specification has a content negotiation mechanism which allows a device to give a simple profile of its capabilities whenever it asks for some information. Even when a server serves the same static content to mobile and fixed systems, Cascading Style Sheets (CSS) allows specific style information to be applied by hand-held clients only, allowing quite different presentations to be displayed in the two cases. These systems, just a few of the technologies which already exist, leaving aside those which could be designed, are much more powerful than a top level domain name.

The various documents about the ".mobi" Top Level Domain talk about not only mobile devices but "mobile users" and "mobile businesses". There is an indication that the mobile technology providers feel that while one is mobile, or when one is catering to a mobile customer, one is special or different. This may in fact be motivated simply by attempts to increase the visibility of the mobile communications supplier's name. It may be connected with a hope by the communication providers to gain some control of over information flow to and from mobile users. This would be detrimental to the open markets enabled by the Internet.

If neither of these motivations are the cause, then perhaps there is an honest belief that being mobile will indeed be best when it is visible to end users. In other words, the mobile communications providers are expecting to declare failure. It is failure when a communications system, in providing connectivity, becomes foremost in the user's perceptions. A travel agent should be a travel business, not a "mobile business". In a reasonable world, the travel agent gets on with selling flights and not worrying about whether a customer is attached by a wire. In a reasonable world, a phone is a phone and the particular electromagnetics used to connect it to another phone are totally uninteresting compared to the fact that a person is connected to another person.

Damage: Loss of Web Functionality

But the point is not that a division into ".mobi" and the ("immobile?") rest of the world is futile, it is that it is harmful.

The Web works by reference. As an information space, it is defined by the relationship between a URI and what one gets on using that URI. The URI is passed around, written, spoken, buried in links, bookmarked, traded while Instant Messaging and through email. People look up URIs in all sorts of conditions.

It is fundamentally useful to be able to quote the URI for some information and then look up that URI in an entirely different context. For example, I may want to look up a restaurant on my laptop, bookmark it, and then, when I only have my phone, check the bookmark to have a look at the evening menu. Or, my travel agent may send me a pointer to my itinerary for a business trip. I may view the itinerary from my office on a large screen and want to see the map, or I may view it at the airport from my phone when all I want is the gate number.

Dividing the Web into information destined for different devices, or different classes of user, or different classes of information, breaks the Web in a fundamental way.

I urge ICANN not to create the ".mobi" top level domain.

Tim Berners-Lee

Cambridge, Massachusetts, 14 May 2004

See also:

[UW]: Berners-Lee, T., Universality of the WWW, Japan prize commemorative lecture, Tokyo, 2004. [slides]

[WTW]: Berners-Lee, T. Weaving the Web, Harper, San Francisco, 1999.

[IJ]: Jacobs, I.: Why Using TLDs for Filtering is Ineffective, Harmful, and Unnecessary Public communication. 2004

Up to Design Issues

Tim BL