Why Using TLDs for Filtering is Ineffective, Harmful, and Unnecessary

There have been proposals to the ICANN to create top-level domains such as ".xxx" (See the ICM Registry fact sheet on their .xxx proposal) or ".adult" for "adult material" and ".mobile" for content that has been tailored to the characteristics of mobile devices (e.g., that have small screens and less memory than a desktop computer). In this paper we survey the reasons why creating new top-level names as a means for creating subdivisions of the Internet is ineffective, harmful, and unnecessary. This paper summarizes some of the issues raised by the creation of content- or device-oriented top-level domains; it provides little detail.

This document represents the opinions of the author, or more precisely, summarizes what the author has understood from discussions on this topic that have taken place as part of the author's work at W3C. This document does not represent the consensus of any W3C group or the W3C as a whole. It is a draft document and may change at any time.

1. TLDs Do Not Allow Effective Filtering

On the surface, the primary reason for creating a top-level domain such as ".xxx" is to allow software to filter content more easily that some might deem offensive to themselves or their children. Top-level domains do not allow effective filtering for the following reasons:

  1. Creating ".xxx" provides no guarantee that all "offensive" material will be confined to these domains. Policing the Internet to ensure that no such material appears outside of ".xxx" is difficult and costly due to complications of International law and the sheer size of the Internet.
  2. Similarly, creating ".mobile" provides no guarantee that all material in the ".mobile" domain will be well-suited for mobile devices. Indeed, it would be highly undesirable to create a top-level domain where only a small set of publishers controlled the content published in those domains.
  3. Perhaps initially, searching and filtering will be improved as the space is filled by content from large mobile providers. Over time, it will become difficult to ensure that deployed content evolves with changes in technology. Ultimately, the utility of a site depends on other factors than the domain name; trust derives from reliability and persistence, which are social issues, not technological issues. ICANN should not mislead the public by setting expectations that become more difficult to satisfy over time and as the amount of content and number of content providers increase. ICANN should not mislead policy markers by suggesting that certain technical approaches may solve social issues.
  4. Arbitrary boundaries undermine persistence over time, as technology and society evolve. A bare ankle one hundred years ago may have ended up classified as .xxx.
  5. Many entities with an established presence in existing domains such as ".com" would be unlikely to "pack up and move" to ".xxx"; this would be disruptive to their business. Instead, they would likely purchase a new domain name in ".xxx" and maintain their ".com" presence (e.g., by redirecting requests to the .com name to .xxx). Filters based on the name alone (as opposed to information available through the transfer protocol) would miss the connection.
  6. Anyone can register a domain name and map it to an arbitrary IP address. Thus, I could register "takethat.xxx" and point it at my competitor's site "teddybears.example.org". This would likely fool filters based on ".xxx" to keep out legitimate requests for information at teddybears.example.org.
  7. The definition of what is offensive obviously differs greatly from country to country, from year to year, and from person to person. If bare ankles are considered obscene in some cultures, but are permitted in photos of Web sites in France selling sandals, then individuals wishing to keep photos of bare ankles out of their home using filtering on ".xxx" are unlikely to succeed. How will sites about safe sex or AIDS be treated? Who will establish what is art and what is pornography?

2. TLDs Created for Filtering are Harmful and Burdensome

  1. Companies protect their trademarked names. IBM is likely to purchase "ibm.xxx" to protect their trademark (not to mention that they would certainly be loathe to allow pornographers to own "ibm.xxx"). The creation of new top-level domains leads to a rush to rent domain names, which is burdensome and imposes extra cost.
  2. Suppose a ".mobile" domain is created for content tailored to mobile devices. Suppose also that ".xxx" is created and ICANN mandates that all sexually explicit material appear in that domain. What does that mean for content providers that wish to provide sexually explicit material that is tailored to mobile phone users? Suppose that China mandates that all businesses put content in the ".cn" domain. What should a Chinese adult shop do? Experience shows that deriving information from content (e.g., google) or metadata is a more effective approach than embedding metadata in names. Strict association of content types with names does not scale.
  3. Device-independent design promotes robust content that may be repurposed in many contexts, some of which the designer may not even have foreseen. For instance, content that does not rely on the capabilities of a particular environment (e.g., a desktop computer with a large color monitor) is more likely to be usable in a variety of other environments, such as on mobile devices, by people with some disabilities, and in environments that may appear in five or ten years. Creating a special area of the Web "just for mobile content" may have the effect of reducing the flexibility of general-purpose content. This will have a negative effect on a number of communities, including users of mobile devices, who may find themselves only able to access a limited portion of what the Web has to offer. A more constructive approach to improving the experience of users with mobile devices is to make all content flexible. Device-independent design saves money for content providers who only have to manage one copy and one URI, instead of multiple sites for different devices. Content designed this way is also generally easier to maintain over time. Users also benefit since more of the Web becomes available to them in more environments.
  4. Attempts to create subdivisions of the Internet based on content may very well lead to privacy violations. It is easy to imagine some governments monitoring users who seek out information in a ".xxx" domain. In the case of ".mobile", I may want some users to know that my Internet-enabled point-of-access is a cell phone, while I don't want others to know; putting metadata into the name limits my flexibility (not to mention limiting number portability).
  5. Adding a top-level domain for each special interest is likely to lead to substantial burden in administration of the DNS, as well as increase the likelihood of error and failure. Using the DNS to create a centralized "yellow pages" of content types is very likely to fail. Indeed, one reason the Web has succeeded is that it does not mandate a single classification: no single classification of content types or device types or other topics will meet the needs of every community. The Web allows each community to maintain a presence in the way best-suited to that community, unconstrained by the needs of other communities.
  6. It is a strength of the Web today that it is possible to reuse much content in a variety of situations. Subdividing the Web not only requires increased effort by content providers (to build many specific Web sites instead of one generally useful Website), it increases the burden on software developers (who may have to implement multiple standards), and inconveniences users (who may not be able to access content because they don't have the "right" piece of hardware, software, or human capability. Attempts to subdivide the Web have failed because ultimately the end user loses (cf. the "browser wars" and the defunct WAP Forum). Part of the value of the Web derives from there being one Web.

3. TLDs Created for Filtering are Unnecessary

  1. RFC 1591 set expectations about how the initial top-level domains were to be used. In particular, ".com" was intended for "commercial entities" and ".org" was intended for "miscellaneous" organizations. In practice, today there is no real distinction between ".com" and ".org" and nobody cares or is inconvenienced by that fact. For instance, ibm.org simply redirects to "ibm.com". Also, when I seek to register a domain name, I am usually given the choice of one or all of "myfav.org", "myfav.net", or "myfav.com". The top-level domain doesn't matter to me; what matters to me (for purposes of branding) is the part that comes before the top-level domain. Other examples: ".tv"
  2. Experience shows that the companies prefer ".com" for their primary top level domain in the market. Some companies in Germany use domain names ending in ".ag", but not as their primary domain name since it is less familiar to people than ".com".
  3. Companies can design their content to work well on multiple devices, including mobile devices. This can be achieved through existing technologies such as style sheets [CSS].
  4. Companies can identify content that is suited for mobile devices using existing protocols, such as Composite Capability/Preference Profiles [CCPP]. If they wish for the word "mobile" to appear in the name, they can already create "mobile.mycompany.com".
  5. Client-side filtering products are widely available.
  6. Companies can use metadata labeling mechanisms more effectively.

4. Who benefits from these proposals

  1. Registrars
  2. In the case of .xxx, pornographers. Professional pornographers benefit from improved branding, likely safe harbor from legal woes, and there is likely to be a decline in "amateur" pornography as some people may not wish to migrate to the new TLD.

5. References

Composite Capability/Preference Profiles (CC/PP): Structure and Vocabularies 1.0, W3C Recommendation, G. Klyne et al., 15 January 2004.
Cascading Style Sheets, Level 2, W3C Recommendation, B. Bos et al., 12 May 1998.
[RFC 1591]
Domain Name System Structure and Delegation, IETF RFC 1591, J. Postel, March 1994.
[RFC 3675]
.sex Considered Dangerous, IETF RFC 3675, D. Eastlake 3rd, Motorola Laboratories February 2004.
New top level domains considered harmful, Tim Berners-Lee essay.

6. Acknowledgments

The author has gathered material from a variety of sources, including discussions within the W3C Team, the W3C Technical Architecture Group (TAG), with Joseph Reagle, with Andrew McGlaughlin, and from RFC3675. The comments sent to the ICANN Forum on the .mobi proposal were also very interesting. Those comments and this summary intersect somewhat, but also raise important points about the proposed organizational structure, DNS behavior, and anti-trust.

Ian Jacobs, W3C

Last modified: $Date: 2000/01/01 00:06:47 $ by $Author: ijacobs $