New top-level domains

The Internet Corporation for Assigned Names and Numbers (ICANN) is in the process of creating who knows how many new TLDs, or top-level domains — the outermost name components of the ubiquitous Internet domain name, e.g., foo.bar.com. Right now ".com", ".net", ".org", etc., are the only choices; ICANN wants to add new ones like ".app" and ".beauty". The New York Times has an article about the matter. (The article, I should add, incorrectly dubs ICANN "Icann". Forget that idiocy: it's ICANN.)

I can't fathom why ICANN has allowed the domain naming scheme to be mauled in this way.

Existing top-level domains like ".com" and ".net" conform to an organizational scheme that goes from general to specific. At the top level, sites are organized into broad categories like ".gov" for U.S. government sites and ".net" for network providers. The administrators of each TLD were then free to organize that TLD any way they chose, and each subdomain was free to organize that subdomain any way it chose. The University of California, Berkeley, for instance, having been granted the "berkeley.edu" subdomain (or simply domain), could then create more specific subdomains like "haas.berkeley edu" (for the school of business) and "kalx.berkeley.edu" (for the campus radio station).

The choice of the original TLDs was arbitrary, perhaps even capricious. Probably no one outside the U.S. government, for instance, is happy that ".gov" is the exclusive province of the United States government. There's definitely room for improvement in the ad hoc organizational scheme embodied by the existing TLDs. As exasperated as I get with the exhaustively detailed standards promulgated by ISO, I suspect only the kind of painstaking, comprehensive approach that ISO takes to its problems would result in a domain naming scheme that would be robust and extensible. (Of course, ISO solutions tend to have the drawback that they're largely incomprehensible to mere mortals and are a royal pain to implement, thus limiting their adoption. ISO's design for computer network protocols, for instance, is elegant and comprehensive, but to my knowledge the only operating systems that attempted to implement it abandoned the effort more than a decade ago in favor of the rough-and-ready TCP/IP stack.)

Nevertheless, as much as change is needed, I don't see how the equally arbitrary (or capricious) TLDs apparently being suggested by applicants to ICANN today are going to help matters. They'll increase "real estate", as one so-called expert remarked in the Times piece, but that's about all — and increasing real estate benefits the applicants, not the Internet as a whole. That's fundamentally wrongheaded. ICANN should be trying to make the Internet better, not catering to the whims of companies whose short-term profit motive is at odds with the long-term stability and usability of the Internet.

As far as I can tell from an admittedly quick and incomplete glossing of ICANN's documentation of its TLDs expansion project, ICANN didn't start by envisioning any kind of architecture for domain classification. Instead, ICANN laid out technical requirements for supporting what looks like an uncoordinated rush to create a TLD landscape without any structuring principle. If a name can be unambiguously converted to bits that all name servers can handle, it's good.

I suppose that ICANN's vision is in line with how typical Web users see the Internet today: hierarchies are inconvenient at best when massive search engines make data available at the click of a mouse. Why bother memorizing "haas.berkeley.edu" when typing "Haas School of Business" in the browser's search field brings up the business school's site as the first hit?

Yet an absence of structure in domain names can lead to unnecessary confusion. A TLD is supposed to be a generic category within which more specific objects can be found. However, there are all kinds of generic categories. ".car", for instance, might be a fine place for a Ford dealership to ensconce its domain name, but what about a bumper-car manufacturer, or a site devoted to Pullman coaches that wants to have "train.car" in its domain name? Or what about ".love", one of the proposed TLDs? How many kinds of love are there? How useful is "love" as a category?

The drive to expand Internet "real estate" has trumped any concern for why the original domain naming scheme was structured as it was (and is). If hierarchical/categorical naming is no longer important, why maintain the fiction of domain names at all? Why not create a different naming scheme that doesn't trade in the trappings or structure of the existing domain names hierarchy?

Part of the appeal of the existing TLDs is that they serve as a rough (admittedly, increasingly rough) guide to the purpose of their subdomains. The existing TLDs don't organize the world well enough, we can all agree. Yet is the answer to abandon all hope of organization? That seems to be what ICANN has decided.