Internet domain names are truly bizarre. There is nothing especially remarkable about them from a technical perspective, but from a social and political perspective they are all sorts of fun. We can have arguments over control of the DNS root, arguments over whether names are property, arguments over innate rights to specific names, arguments over a registrar's right (or lack thereof) to exploit unregistered names for private gain, and many more arguments besides. In this article, I'd like to explore the argument-space rather than defend any particular position in it. In so doing, I hope to illuminate some novel (or under-emphasised) perspectives on the matter.
What are domain names like? How much of that likeness is essential, and how much is an effect of policy? What ought they be like? A recent article in the Journal of Intellectual Property Law & Practice drew parallels between domain names and telephone numbers. More specifically, it suggested that rulings relating to telephone numbers are the most appropriate legal framework for domain names. Consider the following extract from the article.
In some respects, domain names are even more akin to toll-free vanity telephone numbers such as 1-800-HOLIDAY. The analogy is appropriate because a domain name is a mnemonic for the unwieldy initial portion of a numerical address. The mnemonic device is often related to a business name or a trade mark and may be the basis of a substantial advertising investment. Each represents a scarce public resource in that the unique nature of vanity numbers and domain names means that only a limited number of each is available. Finally, the continued right to use a vanity number or a domain name is dependent on the continuation of services from the telephone carrier or domain name registry.
There's a degree of validity to this argument, but it's also easy to construct a counter-argument with just as much validity. One might say that a domain name is, unlike a "vanity" number, not a mnemonic, but a genuine layer of indirection; or that the comparison is dubious because the range of possible domain names is vast compared to that of phone numbers, even if both are scarce to some degree; or that a telephone number is an essential part of telephone service, where a domain name is not essential to all forms of Internet service. These and other objections are reasonable.
More seriously, though, the argument seems to assume that things are the way they are as a matter of immutable necessity, which is largely false. Granted, sheer inertia is likely to maintain the status quo, but there's a difference between inertia and immutable necessity. For example, domain names are, to some degree, a scarce resource, but are they necessarily scarce? My Cornucopia DNS concept demonstrates that there are theoretical ways to eliminate scarcity. If scarcity is not necessary, then we ought first to consider whether scarcity is desirable before formulating legal theories around the principle of scarcity.
In the following paragraphs I will consider a number of such DNS aspects, their implications, and whether the DNS can theoretically be changed with resultant changes to the implications.
One aspect of a domain name is its semantic value. Domain names can, and almost invariably do, contain semantically meaningful elements. It's important to note that such values exist in the mind of the beholder, rather than in the DNS itself: semantic values depend on one's ability to understand the words employed. Universally uniform semantics are not possible without a common universal language. Further, even when language is not a barrier to understanding, there is often a distinction between the semantic value as intended and as exploited. How many registrations in ".tv" actually relate to Tuvalu, for instance?
The semantic value of a name — intended or not — may suggest an owner's identity. This is a decidedly mixed blessing, because although a name like "ibm.com" carries a strong suggestion as to the identity of the owner, nothing in the domain name system itself ensures that the implication is true. Naturally, owners of famous names and brands have sought to register their obvious DNS equivalents as they become aware of the need to do so, but sometimes they are too late. This has given rise to the concept of "cybersquatting", and subsequent attempts at legal remedies such as the UDRP. The remedies have, in turn, had ostensibly unintended consequences, mostly involving the owners of famous names and brands confiscating the domain names of sites which portray them in a negative manner.
Semantically meaningful TLDs such as ".com" and ".edu" were an obvious and reasonable choice when initially selected (see RFC 920, October 1984), but that choice of terms has also caused many problems. That's not to say that ".com" and its ilk were a mistake in and of themselves; the framers merely (and excusably) failed to anticipate the full extent of the external social and political pressures to which the DNS would be subject in the long run. For example, the gTLD semantics were based around the idea that registrations in those domains would be restricted to organisations having a certain minimum number of hosts [RFC 920, p.6], but this foundational principle was eroded by various social and economic factors in the 1990s, starting with the contract between the US Government and NSI in 1993, and culminating in the dot-com bubble.
DNS use in current practice bears little resemblance to the picture painted in RFC 920. Although domain names are used to name hosts in hierarchically structured organisations, as was envisaged, that is not their most prominent function. Most prominently, they associate online services with concrete network addresses. These online services are not always appropriate to the semantics of RFC 920 gTLDs and subsequent additions. CircleID itself is an example of this state of affairs: it is, first and foremost, a website in its own right, not a company with a collection of hosts that require Internet domain names. Every discussion forum, blog, wiki, webcomic, webmail service and collaborative software project finds itself similarly neglected, semantically speaking. Even things that are companies typically require their domain names first and foremost for their website and email services, not for host names.
Despite the conspicuous paradigm shift from "hierarchically structured organisations with named hosts" to "named online services", current trends in DNS naming activity retain both the idea of categorical semantic values and (unsurprisingly, perhaps) the abuse of those values. The ".pro" registry seems to be a case in point. Recent reports also indicate that ISPs in certain countries are intercepting root DNS queries in order to overlay names in local character sets onto the DNS. If they've taken things that far, then they may also have gone so far as to introduce their own localised TLDs with semantic values of their own choosing. So far as I'm aware, my Cornucopia DNS concept is the only approach which advocates a semantically void DNS component; sacrificing some of the benefits of semantics in order to avoid the associated problems.
The article from the Journal of Intellectual Property Law & Practice (cited earlier) claimed, in part, "the continued right to use a vanity number or a domain name is dependent on the continuation of services from the telephone carrier or domain name registry." This is true, but only contingently so. We can quite easily separate the concepts of ownership and listing. Imagine, for a moment, a parallel universe in which domains names are sold via title deed, granting the holder the right to the domain name in perpetuity, but not causing the name to appear in the global DNS. The actual listing of a name would be available to the deed holder for an additional annual fee.
This is very much a thought experiment, because control over any domain ultimately rests on the say-so of the root controller, which is ostensibly ICANN at the moment in most parts of the world. Unless the root controller honours title claims, title doesn't mean anything. Indeed, the concept of "title" must be honoured all the way down the chain of delegation if it is to work. Even so, it's a thought experiment worth conducting, because the possibility of actual domain name ownership has some interesting consequences.
An apparent down-side, for example, is that it might facilitate cybersquatting to some degree. The theory is that a cybersquatter need only pay for the title, not the listing, and the title transfer is a one-time fee. This doesn't necessarily facilitate cybersquatting though, as the title fee (by merit of its nature as a one time fee) may be moderately high — say the equivalent of several years worth of listing fees. I don't know much about the actual economic dynamics of cybersquatting, but I would suppose that the longer a domain is squatted, the less likely it is to sell, so a modest title fee may work against cybersquatting rather than for it. Then again, I expect cybersquatting to be a gamble with long odds, so tweaking the price of a domain one way or the other may not be economically significant.
The mention of cybersquatting raises an interesting point: namely, that the secondary market in domain names operates as though there were such a thing as "title" already, because the current registrant of a domain is nine-tenths the owner anyhow. The lack of genuine ownership creates issues the moment a registration expires, however; a situation which is most disadvantageous to sincere, long-term domain owners (as opposed to squatters, spammers, phishers and their ilk). I'm reminded of Talented Fool's article, Riding on Expiring Domains: Are Registrars Abusing Owners' Rights? If there were a clear distinction between title and listing, this problem would resolve itself: an expiring domain would expire only from the DNS; the registrant's ownership of the domain name itself would not expire. Re-activating the domain would be a simple matter of paying the appropriate listing fee.
Another point of interest is the interaction between special entitlement claims and the concept of ownership. Under the current regime, it's possible to obtain a name by registering it when available, by purchasing it from its current registrant if a deal can be struck, or by disputing entitlement under the UDRP. It's a very strange form of property that allows some unrelated trademark owner to claim entitlement to that property. This harks back to the matter of semantics and the suggestion of owner identity inherent in some names. The weight of argument may be shifted in these disputes if it were recognised that the registrant of a domain holds title to that name rather than merely renting it, just as there is a difference between evicting a tenant and seizing a house from its owner. That doesn't make the problem any less tense, but the root cause of the tension is semantics, not ownership, so I won't discuss it further here.
So why do we have "rental" rather than "ownership" of domain names in the first place? In the earliest days of the DNS, there were no fees associated with listing. Perhaps this is what led to the idea that names are borrowed rather than owned: every name was ultimately on loan from IANA. Perhaps there was, prior to the web and the URL, no compelling reason to own a domain name at all. After all, if your model of the DNS is "hierarchically structured organisations with named hosts", ownership seems like overkill; but if your domain name is stored in URLs, bookmarks and address books all over the network, it can become a thing of some personal value to you, and ownership seems only fair. Perhaps IANA allocation of names was just the natural thing to do, following the model by which IP addresses were allocated. Whatever the case, I suggest that reconsideration is in order.
One last thing to note about ownership is that some might oppose the idea in principle due to scarcity. It's a long-standing principle that scarce resources must be leased from some managing authority, rather than owned outright. Usage of the electromagnetic spectrum is managed by government authorities for this reason. But scarcity is a relative concept, and scarcity of names in the DNS is largely artificial, resulting from a restrictive policy with regards to TLD semantics. Again, I refer to my Cornucopia DNS concept: if it were to be implemented, then I would also suggest that names in that space should be genuinely owned by the purchasing registrant, not merely rented from the registry. The namespace described by the Cornucopia is like a vast wilderness, and there's no reason why people should not be entitled to go forth and stake their claim in it. Regulatory involvement is then limited to managing a claims office for the purpose of ownership dispute resolution.
Whether or not a domain is "property" in a legal sense, it is subject to exclusive control, and may be valuable. The value in a domain name can arise, so far as I can tell, from two distinct sources: intrinsic semantic value, and value added by homesteading. Intrinsic semantic value is readily demonstrated by the "sex.com" fiasco. Apparently there is a significant body of curious people who wonder what is at that address, and go there spontaneously — so much that the address itself is a virtual license to print money just by providing ad space. Contrast this with "ebay.com", which is hardly an address one would investigate spontaneously, but is a well-known and profitable business because of the effort put into building a service on that domain name.
I have selected the term "homesteading" to describe this so that I might draw further parallels between domain name ownership and land ownership, a parallel that I introduced with "title deeds" in the previous section. In the same way that domain names have intrinsic value and added value, land has intrinsic value due to location, and added value due to development. There's a world of difference between someone who manages to stake a desirable claim in a land rush, and someone who manages to add value to an otherwise ordinary claim by developing it. So it is with domain names.
Another thing to note about the value of a domain name is that value (whether intrinsic or added) emerges out of the use — mostly third party use — of the name. The exploitable value of "sex.com" is not so much that it is a coveted generic term in the most popular TLD (although this inflates its resale value), but rather that it attracts spontaneous traffic which can then be converted into advertising revenue. Likewise, if a person builds a rich and complex website at a particular domain name, but nobody visits it, then the domain name is not valuable despite the development. No damage would be done to such a site by relocating it at a different domain. Contrast this with a site that has a good ranking in search engines, and many third party pointers to it. Such a site suffers harm if relocated, as a consequence of loss of traffic. A similar argument can apply to domain names used for other purposes, such as email.
To the extent that value arises from semantics, DNS naming policy can influence it. The theory behind Cornucopia, for example, is to create a very flat landscape where an indefinite number of semantically equivalent domain names are available. This equivalence strips semantically special names of their intrinsic exploitable value (and resale value) by dilution. One of the reasons for such a policy is to eliminate "land rush" behaviour. The problem of "land rush" behaviour is well-known (see, for example, ICANN's draft Strategy: Introduction of New Generic Top-Level Domains, pp.4-5), but the possibility of avoiding it (as opposed to merely coping with it) has been largely overlooked.
The three aspects of domain names considered here overlap in an interesting manner. Semantics can suggest ownership (or at least association), and also convey intrinsic value. This can shape the "land rush" behaviour when a TLD is initially opened, and the disputes that will occur later. Meanwhile, value added by "homesteading" or development lends support to the idea that domain names should be ownable in a manner similar to land title. If the majority of value in a domain name comes from the continued "habitation" of that domain by a particular "occupant", and the sweat of that occupant's brow, then it seems only fair that the party in question should be granted some on-going title to the domain name.
Many aspects of TLD naming and management are traditional rather than essential. It will pay to review these traditions from time to time, to determine whether they still constitute good practice in the contemporary environment. Matters of semantics and ownership rank prominently among these traditions.
Originally at Nutters.org.
By The Famous Brett Watson, Internet Services Engineer
|Cybersquatting||Policy & Regulation|
|DNS Security||Registry Services|
|IP Addressing||White Space|
Minds + Machines