TC: I was talking yesterday to Vint Cerf about alternate roots. He said: "This loss of uniqueness I think would be a disastrous outcome for the Internet's architecture and for that reason I feel strongly we should stick to a single root. Now what I want to make sure that is clear is that I'm saying a single root zone file, which is unambiguous in the pointers that go to the resolver for those root zone entries."

KA: I use the term "competing roots" rather than "alternate roots", the later has a pejorative connotation.

It is impossible for different systems of roots to publish the absolutely and perfectly identical root zone file - if nothing else the "source of authority" and name server records for each system of roots must be adjusted to reflect the actual set of servers that publish that version of the root zone.

It's interesting that Vint has shifted to "single root zone file" rather than "single root". The reason that I say that is that ICANN several years ago published a document ("ICP 3" - http://www.icann.org/icp/icp-3.htm ) that argues against anything except a single root. I.e. it is ICANN's own published position in opposition to multiple systems of roots even if they publish the same basic root zone file.

ICP-3 was nothing more than techno-dogma, then and now.

As for "disasterous outcome" - where is the disaster in letting people chose which internet phone book they want to be listed in and which they want to use? If people want to communicate they will find common naming ground - which is why I say that there will be a natural pressure that drives and awards consistency among naming systems.

Moreover, there is this unspoken point about DNS names - they are not invariant. Domain names change over time, often very quickly. That's why we have stale URL's and dead email addresses. The person who is "fred@aol.com" today might be someone else next week. So the idea that DNS is some sort of permanent unchanging handle is false to the core.

Moreover, on the web URL's (which contain domain names) are hardly fixed and unchanging. When someone in Sweden utters "www.google.com" they get an entirely different set of responses than I do when I utter the same URL here in California. In addition, the presence of other modifiers - source IP address and cookies further enhance the differences between results that arise when different people utter the same domain name.

That's life. Cope with it.

What we have and what most folks fail to recognize is that the domain name system is really only a hinting system. The attempts to characterize it as "authoritative" are misleading. DNS only says "this name maps into this record" (often an A for Address record). DNS is sort of like going to the gas station and asking for directions - the answer you get does not include a guarantee that the person you are looking for actually lives at the place you are asking about.

This reveals a major gap in the architecture of the internet - the concept of mutual end-to-end idenfification and authentication. DNS may help you knock on the door of a remote server, but there is no universally established means for you and the other end to identify one another and to prove that that identity is authentic.

This is why porn providers, for example, like to snap up domain names that are relinquished, so that they can capture the residual traffic that comes knocking the names of the previous web servers.

TC: Would it be feasible to allow alternate roots, with some mechanism to avoid collisions, and then to merge entries in those roots into a single zone file?

Not really feasible without a sizeable change in deployed technology. But then again why should we try to achieve that kind of Procrustean uniformity. The world has not collapsed because the Americans and Europeans each use the word "corn" to name different things. And in the use what you get when you ask for "a cracker" very much depends on the context in which you utter the request.

The context is provided by the communities that chose to abide by a chosen name space.

Competing roots, I assert, will evolve, based on the desire of communities to be able to communicate with other communities, with a strong core of TLDs that are found in virtually all systems of roots. That's just like in a supermarket of today - you will find the same core suite of brands, like Campbells soups - no matter which chain of supermarkets you enter (at least in the US and Canada.)

Trademark law can be a tool by those who provide those top level domains to ensure that .com remains .com, .net remains .net, .org remains .org, etc wherever it is found. That's the purpose of trademark law.

Some competing roots may chose to offer boutique top level domains - like my .ewe. Does the presence of these harm the core TLDs and those who chose to use them to identify their offerings and those who want to find them? No. Just because Fred's Supermarkes offer Slurpy Soup alongside Campbells in no way damages the ability of someone going into Joe's Markets and finding a can of Campbells.

There also remains the residual question - which god made the NTIA/ICANN/Verisisgn root the one holy catholic root? Why should not other offerings be allowed to arise and compete on their own merits?

Had the mentality that Vint expresses about competing roots been the rule of the land in 1974 we would never had had an internet - the nascent idea of packet switching would have been cast aside and condemned as a danger to the universality of connectivity offered by the incumbent telephone companies - IP addresses could have been construed as a clear and present danger to the universal connectivity offered by telephone numbers.

Good thing that that did not happen.

In the best of possible worlds most of the pressure that drives the desire for competing roots would not exist. But ICANN's own action of imposing arbitrary and highly exclusive requirements for those who which even to apply to ICANN for "permission" to operate new TLDs has driven people to frustration and to want an alternative.

ICANN has even mandated a particular business model on TLDs that is arbitrary and capricious - take a look at my note on my .ewe business model to see that alternatives are possible - http://www.cavebear.com/cbblog-archives/000159.html

The most articulate argument against competing root is one by Paul Vixie. He doesn't raise a lot of technical arguments but rather gets right down to the root (pun intended) concern that because competing root systems are likely to evolve to so that they have, to a greater or lesser degree, some TLDs that are not universally offered by all, that some people will find themselves in the position of wanting to utter a domain name only to find that their providers do not offer that TLD. The result would be an inability of a person desiring to communicate being unable to do so (without jumping through some technical hoops that are probably beyond the knowledge and capacity of the general internet user.)

That's a valid concern, but then again, I routinely try my best to deny certain classes of people (such as spammers) from consumating their desire to communicate with me. In other words there is a tension between the social utility of universal reachibility and the personal value of being able to control who intrudes upon my privacy. It is with regard to that issue that the discussion of competing roots deserves to be discussed and not some arbitrary dogmatic assertion that universal naming must absolutely and perpetually trump everything else.

By-the-way, there is a completely distinct mechanism that could supplant the existing root system - That is the idea that everyone runs his/her own root. I've done it; it works. And it has no central point of failure.

How could this happen - well there once was a system called "Grass Roots" in which you and I could select the TLDs we wanted to honor (and chose among those who were fighting over the same name, like .biz) and create our own root zone. It was a cool and easy to use system.

An alternative way is to use peer-peer techniques to move root zone files around - root zone files are tiny, when compressed they are usually about 16K bytes, smaller than many of the cutsey buttons that decorate web pages. I can easily imagine a system in which root and TLD information is wafted around the net, like digitally signed pheromones, that users (or their proxies) can accumulate and use to bootstrap their own root servers.

We have to remember also that ICANN has a strong financial incentive to preserve the status quo. Should competing roots arise ICANN would find it increasingly difficult to feed its ever increasing hunger for millions upon millions of dollars per year.

You should check out the Hush-A-Phone case for an example from the 1950s in which AT&T and the FCC tried to quash a harmless technology through the use of baseless techno-dogmas. That situation has a strong core of simularity to much of what is happening with internet governance today, particularly ICANN's stance on competing roots.