DNS 2.0

by Maria on August 23, 2007

My ICANN colleague, Kieren McCarthy, has written an interesting piece on the ICANN Blog about types of new top level domains (e.g. .com, .info). He dusted off a 1997 proposal to put .firm, .store, .web, .arts, .rec, .info and .nom in the domain name system (DNS).

What strikes me is the taxonomic approach of what we now think of as Web 1.0. The TLDs considered ten years ago were attempts to organise the Internet from the top down by category and generic activity type. If and when a process for approving new TLDs begins next year (it’s subject to a vote by the ICANN Board, probably in October), it won’t yield anything like this organised and thematic approach.

Rather than creating a hierarchy of meaning, we’ll see an explosion of ideas pushing up from below. About the only new TLD proposal we know we’ll get is .berlin, which has put a glint in the eyes of city managers and tourist authorities all over the world. We don’t know which new TLDs will be created, but as Kieren says they’ll probably be things like .blog, .news, .coffee, .google and the like, i.e. services in search of a market and branding efforts by companies, cities and pretty much anything you can think of.

The predominantly English-speaking technical cadre that looked at this issue 10 years ago only came up with one non-English TLD (.nom) which was still pure ASCII text. Today, the global technical community is working hard to smooth the way for internationalised domain names, i.e. names in non-Roman characters.

It’s clear that the Internet will start changing as soon as the new TLDs begin to appear. What’s not as obvious is how ICANN may change. Just as the European Economic Community was fundamentally altered by conceiving and administering the Common Agricultural Policy, ICANN may itself be changed by the new gTLDs programme. The CAP is a bad example substantively, as it was designed to shut competition out. The DNS isn’t a way to organise the world’s information, but is a tool people can use to organise and express themselves. I hope the new gTLDs will give expression and form to communities and interests around the world that use the Internet but don’t yet see themselves in it.



Seth Finkelstein 08.23.07 at 4:22 am

By coincidence, I have a column in the _Guardian_ today about some controversy over the procedures to be used for approving new TLD’s.

“What’s in a name?”
“The domain name system is full of speculators, squatters, and scammers.”

I argue against having a “neutral” stance on monopolistic rent-seeking registries :-).


Maria 08.23.07 at 5:23 am

Thanks, Seth – interesting piece.

Do you know why UK press organisations call us ‘Icann’ instead of ICANN? The BBC news site does it too and it always gets right up my nose.


Adrian 08.23.07 at 6:34 am

Could someone remind me why we need TLDs at all, now that the old Yahoo! categorisation approach (and the original .edu vs .com snobbery) is gone? Presumably they’re not technically necessary–DNS servers could translate “crookedtimber” into the string of numbers that actually matter as easily as they can “crookedtimber.org”.

Site name is obviously useful, and server ID is good too when it’s not just “www”. But all TLDs seem to achieve is to allow duplicate site names–is that the point?


Seth Finkelstein 08.23.07 at 6:55 am

Maria, by coincidence, I asked that very question about the copy-editing of an earlier column which had “ICANN” in the original draft (since I am American). Apparently it’s a style rule:


“Use all caps only if the abbreviation is pronounced as the individual letters; otherwise spell the word out: the BBC, ICI, VAT, but Isa, Nato”

Since it’s pronounced “I can”, rather than “eye-see-ay-en-en”, then it’s “Icann” rather than “ICANN”.


Maria 08.23.07 at 7:42 am

Now that does make sense. I was wondering to myself why we say Nato but not Wto. Thanks, Seth!


Maria 08.23.07 at 7:48 am

Adrian, there’s something to that, and it’s often asserted that with search improving all the time, we don’t need the DNS. I don’t completely buy that, since most search algorithms include domain names in their analysis.

One point of having many TLDs is that it multiplies the number of potential and available names, which is increasingly important as more of the world starts to use the Internet.

Another thing to remember is that domain names aren’t used only for websites, but also for email addressing. I’d hate to think what would happen to an email that was simply addressed ‘Maria Farrell’ and sent off into the ether.


Adrian 08.23.07 at 8:11 am

Maria, thanks. I certainly wouldn’t want to get rid of the DNS altogether. My mind’s not that flexible. :-)

But still, if TLDs (as opposed to domain names generally) are there so there can be a Chinese and an Indian “crookedtimber” I don’t see why they need to be such a fundamental part of the architecture. How can a fixed menu of TLDs give you flexibility you can’t get from the unlimited-length alphanumeric site name? I mean, if crookedtimber.ru wasn’t an option, the site name could always be crookedtimberru or crookedrussiantimber or whatever.


Maria 08.23.07 at 8:40 am

Hi Adrian,
Kieren actually partly address this question in the comments on his post; http://blog.icann.org/?p=183. Name exhaustion is becoming a problem, especially in TLDs like .com where pay per click advertising drives registration of vast number of names, locking them out for other uses.

I see the main virtue of TLD expansion as creating a rallying point for communities that don’t have an online identity, or for niche users. So that’s more about creating new use TLDs, rather than expanding the namespace to accommodate the growth in current use patterns. But that’s just my personal view.



Maria 08.23.07 at 8:42 am

Seth, you inspired me to finally break the seal and blog on the ICANN staff blog; http://blog.icann.org/?p=185


Adrian 08.23.07 at 9:32 am

I guess the common thread is that TLDs allow some measure of registrar control over the use of certain strings in domain names. This allows the TLD to categorize sites–which we all now think is outmoded–and to signal something about the site’s provenance (in the .edu, .mil etc domains). Which is fine, though I s’pose you could still have privileged strings even if the end of domain names was otherwise unregulated. (It also gives the registrars something to do besides scan for rude words.)

Kieran’s point is, in effect, that this control over terminal strings means no one can currently have a name ending in “.new” because .new hasn’t been created (allowed) yet. So “first-choice.new” hasn’t yet been snaffled up by a PPC farmer or some other squatter.

But what happens when .new is allowed? The same squatters will send their bots to the .new registrars. And very soon you need .newer, and so on.

The “name-exhaustion” problem is different in this respect from the exhaustion of IPv4 numbers (or seven-digit phone numbers), which really can be resolved by making more numbers available for new users and devices.


Pete 08.23.07 at 9:50 am

“Name exhaustion is becoming a problem, especially in TLDs like .com where pay per click advertising drives registration of vast number of names, locking them out for other uses.”

This is ridiculously wasteful, surely? Perhaps what’s needed is a .unique domain, in which no one person or organisation is allowed more than one registration.


john b 08.23.07 at 12:38 pm

“The predominantly English-speaking technical cadre that looked at this issue 10 years ago only came up with one non-English TLD (.nom) which was still pure ASCII text. Today, the global technical community is working hard to smooth the way for internationalised domain names, i.e. names in non-Roman characters.”

great: even more scope for incompatability, impenetrability and other hilarious things. if only everywhere with a machine-unfriendly writing system had had an Ataturk…


Alex 08.23.07 at 1:11 pm

Regarding tlds, this might be a good moment to recall that one of the points of Python Zen is “Namespaces are a honking great idea; let’s have more of them.”

Every new tld is a brand new namespace, and the structure of the DNS means that (unlike IP routing tables) it’s easier to scale to fit a bigger set of lookup tables.


robertdfeinman 08.23.07 at 2:17 pm

At least 95% of people reaching my web site do so either via a search engine or by clicking on a link on some other site. I imagine this is typical.

The idea that people will type in a site name seems a remnant of an earlier era and really needs to be rethought. Bookmarks and other personal convenience features also make direct typing less likely.

Haven’t we gotten along with a numeric system for phone number for over 100 years? The needed access tools like phone books and directory assistance evolved to solve the problems. The internet is no different.


Tom 08.23.07 at 4:38 pm


There are good technical reasons for using domain names rather than relying on search + IP addresses (and remember, not all of the internet’s users are people sitting at web browsers). A domain name functions as an alias, allowing a site to avoid having its name tied to the physical infrastructure supporting it while still allowing the underlying IP address to be organized in hierarchical blocks. Keeping IPs organized in continuous blocks (rather than offering some version of telephone number portability) makes routing easier to implement efficiently.


rupes 08.23.07 at 4:38 pm

Talking of domains there is one amusing anomaly…

As well as categories (.com, .edu) of course, domains can be countries: .uk, .de, .fr etc
These are two letter based on ISO country codes

Of these Tuvalu apparently has a revenue stream from .tv

The amusing anomaly is .cat for Catalunya

1) It is a country code, but three letters not ISO two letters

2) It is a country code for a place that is not a country.

Or rather, where inhabitantswould like to be a recognised country, but currently are not.

Or even more convolutedly, shows up anomalies of “country” as a description.

After all, ir Catalunya gets recognition by ICANN then why not Scotland, Corsica, any-number-of-disputed places in Balkans & Central Asia etc

Even more intriguing, within Spain, Basque contry has more autonomy than Catalunya, but not in this respect :)

I have no idea quite *how* they managed to persuade ICANN to agree they are a country…


Melinda 08.23.07 at 4:41 pm

The point about the scalability question is apt; the DNS is hierarchical in part because it makes delegation easier and in part because it just plain scales better – flat namespaces can be extremely difficult to manage.

A bunch of years ago I was involved in ETSI and ITU-T discussions about allocating an E.164 prefix to VoIP (this was before ENUM, obviously). A couple of European countries that had recently privatized their PTTs and a bunch of African countries were very opposed to it, largely because phone numbers are property and have value. Personally, the extent to which that value tends to lead to misbehavior (domain name squatting, for example) is pretty much the extent to which I support undermining that value.

Unfortunately economics sez that proliferating TLDs will tend to lower the cost of registering in each of them and the whole thing might well end up being a wash.


mollymooly 08.23.07 at 11:05 pm

@16: According to Wikipedia, .cat is a “generic top-level domain” rather than a “country code top-level domain”. It is for the Catalan language and culture (which are international) rather than the particular Spanish region where the culture predominates.


john b 08.24.07 at 9:24 am

See also: Catalans rugby club, who are French.

Comments on this entry are closed.