Due to a disk crash and backup failure, this site has been restored from an old backup with a number of more recent articles missing. The missing site content is being restored as time permits. We apologise for any inconvenience.

Cornucopia: A Radically Different Approach to TLDs

or "What if every domain name you wanted was available?"

A Constrained Rant by The Famous Brett Watson, 01-Dec-2004.


Much of the discussion about proposed TLDs centres around domain names as a form of classification: ".mobi" for mobile device content, ".kids" for child-safe content, language codes for language-specific content, ".museum" for museum-related entities, and so on. Notoriously little activity has been forthcoming in actually implementing these proposals, and the select few that have been allowed out into the world are, shall we say, a tad arbitrary.

I'd like to engage in a little thought experiment where we abandon the "few TLDs with carefully chosen meanings" paradigm, and instead consider the benefits of a cornucopia of completely meaningless TLDs. This isn't a proposal — yet — but rather an idea which I think deserves an airing. You may think it intriguing, or possibly the worst idea in the history of bad ideas: any well-reasoned feedback is welcome. If your immediate reaction is "why would anyone want a meaningless TLD?", then please bear with me — I'll get to that point shortly.

A Sketch of the Idea

Imagine that a large slab of root namespace has been set aside for the cornucopia: just as all two-letter TLDs are reserved for country codes (whether currently allocated to a country or not), let us imagine that all "one letter followed by at least two digits" TLDs are reserved for the cornucopia. The number of possible TLDs in the cornucopia is therefore so large as to be inexhaustible. We will, however, only populate this namespace on demand, starting with the shortest names: the "alpha-digit-digit" names (2600 distinct TLDs).

The existing registry/registrar/registrant arrangement would be maintained, with some subtle differences. Rather than specifying the full domain name, the registrant specifies only the second level part. Thus, the registrant says to the registrar, "please give me a 'foo' domain from the cornucopia". The registrar passes that message on to the cornucopia registry, and is allocated such a domain name at random from the available pool. It may be "foo.h12", or "foo.p45", or any of the other available "foo" domains in the cornucopia.

If there are no such domains left in the cornucopia, then it's time to open up another block of namespace. If the registrant is unhappy with the allocation (because "h12" is their unlucky number, or something) then the only option is to discard the old name and register a new one. This reinforces the meaninglessness of the TLDs by making it expensive to fish for those few that might be meaningful in specific contexts, like "b52".

Would Anyone Buy It?

So why would anyone want a meaningless TLD? The first and foremost reason: generics! Generic terms are highly coveted in ".com" space, and it's safe to say that they are all taken. In fact, they're probably all taken in all the available TLDs that allow registration of generic terms. The cornucopia allows everyone to register their favourite generic term, since there is plenty for all. The meaningfulness of the second-level name can make up for the meaningless TLD to some degree. A less compelling but equally valid reason to want a meaningless TLD is the case where none of the available meaningful TLDs convey the meaning you want: better to have no meaning than the wrong meaning.

In most cases I would expect people to prefer one of the traditional TLDs over a random cornucopia name, but the cornucopia name will be available for registration, whereas the traditional TLDs are a hit and miss affair, with more miss than hit. The existing scarcity of "good" domain names forces most would-be registrants into compromise, and the cornucopia offers another path of compromise.

Technical Considerations

The proposed initial 2600 domains would increase the number of domain names in the root by an order of magnitude. This is no small undertaking, but it could be broken down into arbitrarily small chunks for the sake of manageability. Provisioning of names could then be ramped up as confidence in the process increased, and as demand required.

The potential number of TLDs in the cornucopia is very large, of course, but it is unlikely to grow the root to anywhere near the size of most second-level domains. According to Verisign's "Registry Operator's Monthly Report" for July 2004, there were over thirty million registered names in the ".com" zone and over five million registered names in the ".net" zone at July 31, 2004. The cornucopia only needs to have about as many TLDs as the most popular registered term, and it seems unlikely that this will approach — let alone exceed — the size of existing viable zones.

The cornucopia also lends itself to a much more balanced distribution of names in the DNS. In contrast to the existing tiny handful of names at the root, followed by more than thirty million in ".com", the random allocation of cornucopia names into second level domains will create a much more even distribution between the root and its immediate delegates. Given that the existing massively unbalanced system works, it's not clear how much it really matters, but it seems more technically elegant.

Economic Considerations

Cornucopia seems very attractive from an economic perspective because it eliminates one form of artificial scarcity. I believe it does so in a relatively gentle way: the introduction of cornucopia would reduce the number of new registrations in ".com" (it's bound to take market share from somewhere), but it would not undermine the intrinsic value of the meaningful TLDs, causing a market crisis in the process. A "readjustment", perhaps, but not a crisis. Then again, I'm not an economist. Some educated commentary would be appreciated on this point. What economic effects can be anticipated from cornucopia?

A past CircleID article (Part I, Part II, Part III) talked about auctions as a means for letting the market decide which TLDs were more valuable, but this doesn't apply in the plentiful and undifferentiated namespace offered by cornucopia. I would expect registration prices to fall towards the actual cost of provision, with very little possibility of a secondary market (in stark contrast to the traditional namespace).

Legal Considerations

Some rich pickings arise when legalities are considered: think Trademarks, Dispute Resolution, and Cybersquatting. Tackling that last point first, how does one "cybersquat" an indefinitely large namespace? It's not feasible. No matter how many times you register "X" in cornucopia, the next registrant should be able to do the same. Cornucopia should render cybersquatting impossible by design. That in itself should put an end to dispute resolution, right? After all, dispute resolution is only an issue where ownership is exclusive, and two (or more) parties are arguing about it. Sadly, my observation of the legal process to date gives me cause to doubt that this logic will be adopted voluntarily, largely because of Trademarks and the Deplorable Doctrine of Dilution.

Trademarks may well remain a problem, but this has more to do with the intrinsic problems of trademark law than with cornucopia as an idea. If it becomes impossible to pre-emptively register every possible domain "X", where "X" is the trademark of a company with a large and savage legal department, what will said legal department do? I'm going to assume they'll use every available means to attack other registrants of "X", and I can only hope that cornucopia will act as a pacifying influence for the following reasons.

The lack of a theoretical need for "dispute resolution" (due to unlimited supply and non-exclusivity) should be enough to persuade ICANN that cornucopia does not require a specific dispute resolution process: both parties can register "X" (within cornucopia) so far as the Internet is concerned, and if some law says otherwise, then take it to the appropriate court. On top of that, "X.com" can be understood to mean "X, the company", and "X.de" can be understood to mean "X, Germany", but what do we understand of "X.j31"? The TLD component is quite specifically meaningless, so we can not make the term as a whole mean something. I hope we can argue the moderate position that usage of "X" in a domain must lend itself to interpretation as a trademark before it infringes, and a cornucopia domain containing "X" would not in and of itself be sufficient evidence of infringement (although a particular use of the domain name might).

Note that there's a distinction to be made here between obviously malicious registrations and simple name collisions. Phishers and other malefactors maliciously register names that are similar to some victim's well-known mark in order to defraud or otherwise profit dishonestly. This already happens, and cornucopia may offer some additional scope for it, but it won't change the nature of the problem or its remedies. Addressing this kind of criminal activity is not (and ought not to be) within the realm of trademark disputes.

In short, I believe the cornucopia proposal offers hope that we might re-open some of the middle ground between rampant cybersquatting and rampaging hordes of trademark lawyers, as well as pushing back the legal tussles to within national boundaries, rather than to international organisations of dubious representative legitimacy.

Social Considerations

Internationalisation is perhaps the biggest buzzword in DNS circles, so far as social issues go, and it's a complex and multifaceted problem. Cornucopia doesn't make any direct positive contribution to internationalisation as I see it, but it does "get out of the way" a little. For starters, the TLDs in cornucopia are meaningless and randomly assigned, so no language gains any semantic benefit from them. It's non-preferential in that sense, at least. All questions of language representation within the namespace are reduced to technical questions, since there are no policy arguments as to whether a particular name should be allowed or not. If it's a valid DNS name, it can be registered — end of discussion.

In the more general sense of "what people want from a domain name" (see also Brad Templeton's "Goals for a domain name system" and CircleID article An Economic Analysis of Domain Name Policy - Part I), the cornucopia offers necessarily mixed results. Necessarily, because many of the things that people want are mutually incompatible. "Memorability" is one of the usually desired qualities, and cornucopia TLDs are meaningless (which hinders) but short (which helps). The second level has an endless supply of any word you want, which can potentially aid memorability, convey meaning, and provide an outlet for personal expression. Cornucopia names are not guessable: you'd have to take a "search engine" approach to finding an already-existing domain. Lack of guessability is a disadvantage to business entities (who generally covet a readily-guessable presence in ".com"), but not to John Smith, who just wants a personal domain name and is resigned to it being unguessable one way or another, since all the obvious ones are taken by other John Smiths.

Bear also in mind the social aspects of the economic and legal considerations, which I hope will result in cheaper costs and reduced risk of domain registrations being overturned by marauding bands of lawyers. Cornucopia offers tremendous potential for a "lifetime domain" because there's very little reason to want someone's already-registered name rather than a fresh one. Contrast this with the state of affairs in ".com", where covetousness is the norm.

Variations on the Idea

In the above scenario, I propose an initial 2600 meaningless TLDs, with space reserved for more as the need arises. An alternative to consider is this: if a particular word is so popular that all 2600 instances are taken, then promote it to a TLD of its own. Existing cornucopia registrations using that term could be grandfathered into the new TLD by reversing the terms, such that registrants of "X.Y" were also granted "Y.X". The rationale here is to reduce root zone growth: the really popular words can become zones in their own right and grow those zones as needs be without expanding the root. Deeper analysis would be required to determine whether this idea has sufficient merit to overcome its obvious difficulties.

More trivially, there may be an argument for omitting certain letters from the cornucopia TLDs, simply because the letter "l" and the number "1" are so visually similar, or because "n" and "m" sound similar. We can afford this kind of tweaking, because it's all syntax for the sake of convenience, and no semantics.

Concluding Remarks

Cornucopia certainly changes the dynamics and politics of domain names. Technical issues aside, the sheer weirdness of it may prevent it ever being considered seriously. On the other hand, the prospect of everyone's favourite words becoming available as domain names is pretty attractive. It is, perhaps, the proposal that's come closest to reaching the oft-neglected ideal of "a domain for every person, for life". Is it workable, or can it be made to be workable? Would it be a good thing if it existed? I'm inclined to think so, but I'm even more sure that the idea needs ample discussion before it can go anywhere in practice, so please, let the discussion commence.

Nutters.org Author: The Famous Brett Watson
Date: 2004-12-01
Public Domain: the author waives copyright on this document. Other sources (if any) are quoted with permission or on the principle of "fair dealing" and retain their original copyrights.