The dotDNS proposal is really about binding -- how you connect things.
One consequence of working with computers is that we take general philosophical concepts, such as binding or tying things together, and use them as basic tools. It is far more efficient to just change a pointer than moving a lot of bits or do complex computation.
This is something we do all the time. For example, when we move, we can leave a forwarding address with the post office instead of having to go back to the old house to pick up our mail. What is different is that computers have forced us to explore the concepts and understand them.
We can see the evolution of the concept in thinking about phone numbers. When the phone system was first automated the dial sent pulses that moved a relay. Each pulse moved it one step. After a pause that relay would lock into that position and the next digits would be sent to the 8th relay in the next group. A phone number was really the description of a path through a series of relays. In a modern phone system the number is just a handle that is used to look up the actual path in a table. There are already a number of services that let people change the mapping themselves.
I recently had to replace a credit card and this means I have to now go and change the numbers I've set for automatic payments. It would be much easier if each of the payment entries would have a unique identifier that would be associated with my credit card account and not the particular card which can change. I have the same problem if I lose a Visa or MasterCard since my wife's card has the same number. American Express issues a different number for each card which makes a lot more sense. In ancient computers (the 1960's) the extra lookup to translate a card number into an account number might have been considered difficult but such efficiencies are no longer justified, especially when the rigid bindings create problems. Losing a credit card may mean creating a whole new account and replacing a set of cards rather than requiring little more than the replacement of the card.
Email addresses illustrate the problem people have in understanding the concepts of naming and binding. The use of the term "address" builds upon the idea of the post office address. But it's a naive version of the postal address. While the post office is very flexible and allows you to use multiple names and descriptions for one delivery point and then lets you keep the envelope so you can do the final delivery, email systems are often very restrictive and limit you to just one address per mailbox. And if you change your name (as women often do when getting married) you must change your email address! This is a strange restriction since email addresses are usually listed in directories and the address itself need be nothing more than an arbitrary token. After all, how many people can guess the email address for Joan Smith anyway?
The Domain Naming System was created by people who did understand binding very well. The ability to use English (for example) words seemed harmless enough, especially within a local work group. After twenty years that expedience has become a monster. The fact that such names are used commercially requires that we face up to issues such as trademark which is really about human interpretation and not about technical mechanisms.
While it is too late to undo what we have, it is straightforward to insert an intermediate layer below the .COM names and thus separate the meaning from the abstract pointer.
The real problem is that the concept of binding is understood implicitly but we need to make the concept explicit in order to avoid confusion.
The concept itself is not new. In fact, a thousand years ago the Domesday Book in England did just that but William the Conqueror didn't have to deal with WIPO and could just declare surnames to be abstract identifiers. To this day we readily accept that a Miller doesn't necessarily mill anything. Women, though, didn't need their own identities and needed only to be listed along with other property.
dotDNS is simply the modern version of the Domesday book but today we must use abstract identifiers rather than words and we can do so since computers provide us the tools to maintain the bindings.
As to ICANN, those who care about the meaning of names should shift their focus to WIPO and other organizations that concern themselves with meaning. .COM names don't really matter any more than other trademarks and, I expect, that only a few will have staying power and most people will just lookup names. They will store the result of such searches in their computers address book or use them as internal linkages.
The details of how this plays are far less important than simply recognizing that binding is a key concept and giving us the ability to build on it.