The Tragedy of the .COMs
The problem with .com and the DNS. Originally posted on SATN14-Jan-2002

Please send comments and feedback to Bob Frankston.

Note that this was originally published on my blog but because it's length I have moved the essay here.

I’m glad to see more emphasis on the fact that “.com” names are nothing more than a lousy keyword system. But unlike most such bad designs, this one is foisted off as fundamental technology and does real damage by infesting the Internet.

In medieval England, as with other countries, there was a problem in that people’s names couldn’t serve as reliable “handles” for purposes of ownership and, more important, taxation. The Domesday (or Doomsday) book established stable names so that if you moved or changed professions you still kept the same name. Thus Surly Miller would become a nice blacksmith but would keep the same name. Imagine that you were Dan of the Mercury. Of course one shouldn’t carry a radical change too far. Imagine how confusing it would be for women to have their own names rather than just being considered as part of their husband’s holdings.

A thousand years later few people still understand the concept of handles that are not overloaded with descriptions and other implicit agendas. The DNS serves are important purpose in creating stable handles that can be used to reference a set of records which include the “A” record that translates the name into a set of IP addresses and “MX” records that are used for email delivery. The DNS, while not perfect, works quite well. The problem is not the DNS itself but rather it is in overloading it with a stupid keyword system. The idea that one could name one’s lab machines “Saturn”, “Jupiter” and “Uranus” is fine. Humans find such handles useful even though as one swap boards around it sometimes becomes difficult to know what pieces retain the name and the seventh planet provides humor for the immature. And, in the early days this system worked for Universities and the small number of early participants though even then there were some questions like whether “Miami” was the school in Ohio or Florida. Letting each community like “.edu”, “.gov”, “.mil”, and “.com” was an attempt to reduce the problem as long there were few participants. Oh yeah, there are other countries, so we gave them worry about their own rules.

Unfortunately this made the problem management and created worse problems by raising the ante on the names and requiring the creation of rules about the use of the names that added an agenda with terms of service. Thus you had to be an educational institution (by someone’s measure) to become a “.edu”. An extreme form would’ve been a “.xxx” domain that presumed that there was a clear, unambiguous, unchanging and authoritative definition of pornography.

Authoritative? That a problem in its own right – to use the example of – who is vouching that this is the real “via”. It even presumes that we’ve nailed down the meaning of “identity” – something that is akin to being able to translate all poetry from one language to another. The classic example is the “vodka is strong but the meat is rotten”. That’s a classic example of the early attempts at language translation for “the spirit is willing but the flesh is weak. My point is that this is an arguable translation since the idiomatic translation is not necessarily the correct one. It is not just naïve and foolish to expect that the DNS can provide an unambiguous translation of a name, it is irresponsible to pretend that we can have an algebra for trust.

Even worse, we have the pre-medieval concept of a magic name that can be used as an instrument of control. If you offend anyone your name will be taken away. Thus if the RIAA doesn’t like the music you have, you lose your “.com” name. This doesn’t really take you off the net but it does provides an opportunity for mischief on the part of those who are trying to prevent what they see as mischief.

The use of names for commercial purposes has resulted in an artificial scarcity of “good” names and thus a high annual fee for the use of the names – you can’t even own your name! This means that the names must go bad after a period thus guaranteeing that the Internet must unravel! And it also means that names get reused and the links you have on your school site “go porn” without notice. This has made schools fearful of links. We see the same problem with phone numbers in real stories in which a rape hotline becomes a sex site. Remember that the phone numbering system was created by Strowger when he discovered that calls to his funeral home were being redirect to his competitor by a phone operator (person who connect the calls).

Somehow we’ve slipped down a slope from a simple mechanism for stable handles and now find ourselves in the company of the saber tooth tigers in the La Brea tar pits.

One of the weird aspects of this “.com” debate is that we have clear examples of commercial mechanisms and their limits. The trademark system demonstrates the complexities and compromises necessary for even a simple system of identifiers. Unlike the DNS, near counts. So we can’t have a Rolleks watch. But we do make distinctions between industries and allow Cadillac to be the name of a car, a dog food and a show polish. Names can be reused geographically also.

This is in sharp contrast with the “.com” system in which a simple typo sends one to an entirely different site – usually a porn site. The use of “.com” names is a major factor that makes it seem like the delivery of porn (or to be precise, the delivery of eyeballs to porn sites). I’m not passing judgment on content; I’m just pointing out that human (or “wet space”) systems take such human factors into account whereas a plumbing system like the DNS doesn’t.

And it shouldn’t have to. As I pointed out, the problem is not the DNS, the problem is the misuse of this basic mechanism as if it were also a social mechanism. It is as if we confused the word “true” in mathematics with the word “true” in English and expect that “no no!” meant yes. Sometimes it does as in “there is not no exit – just follow the egress sign”. But usually it is simply emphatic. Flammable and inflammable are synonyms while cleave and cleave are antonyms. Sorry, that’s a bit aside but demonstrates the complexities and richness that the DNS cannot and shouldn’t try to capture.

So, with apologies to Laurel and Hardy (for those who remember ancient comedians), “another fine mess you’ve gotten me into”. One can understand how we started out with a simple mechanism that solved a real problem – it was too difficult to maintain the “host” table that mapped names to IP addresses. The DNS automated the process and allowed the management of the names to be distributed to zones that defined subtrees of the naming system.

In the earliest days one could get away with keeping a list of IP addresses on a piece of paper, the DNS worked with a small number of keywords that worked in lieu of a directory system. But this mechanism cannot scale. I’d’ve preferred to say couldn’t scale but we are still trying to scale it up to the “real world”.

As Gillmor’s article reminds us, people know it doesn’t work and are learning that they better not rely on it. But societies have a way of maintaining dissonance – even when something doesn’t work we maintain the myth it does. How many people are still afraid to swim for an hour after eating a peanut or sterilize baby’s bottles while letting them crawl on the floor putting everything in their mouths?

Even among those who know that there is something wrong with the DNS, there is a fear that it can’t be changed. This is troubling since we know the systems doesn’t work and can’t work so why pander to impossible expectations and deny us the ability to have stable linkages and a marketplace for better solutions for how to translate a description of what we want (at least, what we want in terms of linkages to sites on the Internet)?

The good news is that the problem is very tractable! Once we recognize that people already use search engines in preference to the DNS it becomes apparent that the DNS is not the keyword system of choice and it is only going to become worse as we create more TLDs (Top Level Domains) such as INFO and NAME and then repurpose others because initial like TV and TO happen to have cute meanings in some languages.

The bold step is to remember the Domesday book and sever the linkages between wet space (human) use of names and the use of names for stable linkages. To avoid confusion I’ll refer to the latter as a “handle” though even there “handle” is part of CB lingo which itself borrowed from the world of ham (amateur) radio. Clearly naming cannot be defined in an RFC (Request For Comments – an Internet design memo).

The handles in the DNS should be devoid of semantics. They have no intrinsic meaning and only serve as a pointer to a site. (Technically a pointer to the records that include a pointer to the site). Just like we have many ways to lookup a phone number using descriptions, names or cards, we can have a variety of ways to go from a description such as “Movie showing down the street” to a stable pointer to the corresponding site.

How do we get there from here? The first step is to create a new (and final?) TLD called, perhaps “TLD”. It hosts handles without semantics – perhaps just numbers. And since there is no longer anything magic about a single “.”, the generation of names can be delegated to avoid too big of a load on a single registry. The actual assignment of handles is no more complex than handing out the next available number to people online at a bakery. OK, it’s a little more complicated because rather than just handing out successive numbers, these would have check digits and other mechanisms to assure that most typing errors wouldn’t result in pointing to the wrong site.

But people would rarely type in the names themselves. They would follow descriptions and the keep the handles in their documents or local address books.

Without the deadening effect of the DNS’s claim to be the authoritative keyword system, I would expect a flourishing marketplace in solutions that are far better than the DNS for finding things and for managing ones links.

Since the handles would never be reused the net itself can become much more stable and since the links themselves will not be repurposed simply because one forgot to renew ones lease on the name, it becomes much safer to link to other sites. Sites may disappear but it is far better to have the link fail than change meaning.

We also get past the paranoid view that ICANN is a conspiracy to keep the DNS from working right. The problem is that it can’t do the impossible. Unfortunately, it is not doing the one thing that is possible which is to moot the current “.com” naming system and shift us away from the fantasy that there is a single source of unambiguous and authoritative meaning. Rather than attempting the impossible, ICANN should return to its roots in Jon Postel’s office and do little more than hand out the next available handle to each comer.

This is all too obvious. Yet I continue to find myself getting hoarse saying that time is neigh!